KR101039204B1 - A motion vector prediction method in multiview video coding and a method and apparatus for encoding / decoding a multiview image using the same - Google Patents

Abstract

The present invention provides a method of predicting a motion vector in multiview video coding and efficiently encoding a multiview video using multi-view video coding by efficiently obtaining a motion vector prediction value (MV _P ) in multiview video coding. / Decryption method and apparatus. In the motion vector prediction method according to an embodiment of the present invention, together with the motion vector MV _N of the neighboring blocks of the current block, the motion vector of the time-direction corresponding block of the temporal reference image in the same view as the view to which the current image belongs. MV _T ) and / or a motion vector prediction value MV _P is obtained using one or more motion vectors selected from the motion vectors MV _V of the view direction corresponding blocks of the view direction reference image in the same time zone as the current time zone. According to the present invention, since the motion vector prediction value (MV _P ) is obtained in consideration of the characteristics of the image referred to by the current block, that is, referring to the time direction reference image or the view direction reference image, the prediction encoding is performed by performing more accurate prediction. The efficiency can be improved.

MPEG, scalable video coding, multiview video coding, temporal predictive coding

Description

Method for predicting a motion vector in multi-view video coding and encoding / decoding method and apparatus of multi-view video using the predicting method}

1 is a diagram illustrating an example of a prediction structure in multiview video coding.

FIG. 2 is a diagram for explaining an example of a method for obtaining a motion vector prediction value MV _p defined in H.264 / AVC.

FIG. 3 is a diagram for explaining another example of a method for obtaining a motion vector prediction value MV _p defined in H.264 / AVC.

4 is a view for explaining another example of a method for obtaining a motion vector predicted value MV _p defined in H.264 / AVC.

5A and 5B are diagrams for explaining another example of a method of obtaining a motion vector prediction value MV _p defined in H.264 / AVC. FIG. 5A is a case in which the current block has an 8 × 16 pixel size. And FIG. 5B is a case where the current block has a size of 16x8 pixels.

6A, 6B and 6C show an image " Ballroom " which is one of the test image sequences in the standardization process of multi-view video coding, wherein the size of the image group in the time direction is 12 and defined in H.264 / AVC. 6A shows a part of the entire prediction structure, and FIG. 6B shows an image B corresponding to the time T0 of the S1 view in the prediction structure shown in FIG. 6A. _The motion information of some macroblocks of ₁ is enlarged and displayed, and FIG. 6C is an enlarged display of motion information of some macroblocks of the image b ₄ corresponding to the T1 time of the S1 view in the prediction structure shown in FIG. 6A. It is.

7 is a diagram showing the positions of blocks that can be used to obtain a motion vector prediction value MV _p according to the present invention.

8 is a diagram illustrating another example of a prediction structure in multiview video coding.

FIG. 9 illustrates a combination of FIGS. 7 and 8.

10 is a flowchart illustrating an example of a method of obtaining a motion vector prediction value MV _P when a reference picture of a current block is a view direction reference picture.

FIG. 11 is a flowchart illustrating an example of a method of obtaining a motion vector prediction value MV _P when a reference picture of a current block is a temporal reference picture.

12 is a case to be divided into a number of partitioned blocks block used to obtain the motion vector prediction value (MV _P), for obtaining the motion vector prediction value (MV _P) is a view diagrammatically illustrating the position of the partition blocks actually used .

FIG. 13 is a flowchart illustrating a method of obtaining a motion vector prediction value MV _P through a median operation on N or less motion vectors according to an embodiment of the present invention.

14 is a flowchart illustrating a process of obtaining a motion vector prediction value MV _p according to another embodiment of the present invention.

15 is a block diagram illustrating a configuration of an encoding apparatus using a motion vector prediction method, according to an embodiment of the present invention.

FIG. 16 is a block diagram illustrating an example of a configuration of the use selector 150 of the encoding apparatus of FIG. 15.

17 is a block diagram illustrating an example of a configuration of a motion vector predictor 160 of the encoding apparatus of FIG. 15.

18 is a block diagram showing a configuration of a decoding apparatus using a motion vector prediction method according to an embodiment of the present invention.

The present invention relates to multi-view video coding (MVC), and more particularly, to a method of predicting a motion vector (MV) in encoding / decoding a multiview image and a multiview image using the same. The present invention relates to a method and an apparatus for encoding / decoding.

Multi-view video coding (MVC) processes video sequences of each view obtained from a plurality of cameras. These plurality of cameras are arranged to be somewhat different in distance and / or in opposite directions with respect to the same object. The multi-view image obtained through the plurality of cameras has a characteristic that the image of each view has a high correlation while the degree of light reflection on the surface of the same object is slightly different depending on the direction. Due to the former characteristics of the multiview image, there is a difference in brightness and color difference between the images of each view of the same object. Therefore, in order to improve coding efficiency in multiview video coding, the characteristics of the former of the multiview image are considered. shall. In addition, by using the latter characteristic of having a high correlation between images of each view, the coding efficiency of a multiview image can be improved.

In multi-view video coding, which is currently being actively standardized, based on H.264 / MPEG-4 part 10 Advanced Video Coding (hereinafter referred to as H.264 / AVC), which is an international standard for video coding, In consideration of the two characteristics of the multi-view image described above, a method of improving encoding efficiency is sought. For example, hierarchical B, a method performed to support temporal scalability in Joint Scalable Video Coding (JSVC) as defined in H.264 / AVC. The coding process is also applied to intra-view predictive coding in multiview video coding. In multi-view video coding, inter-view prediction is also performed in parallel to improve the coding efficiency.

1 is a diagram illustrating an example of a prediction structure showing an inter-view prediction and an inter-view prediction relationship in multi-view video coding that is currently being standardized. FIG. 1 illustrates an example of eight views, in which a size of a group of pictures (GOP) in a time direction is eight. 1, S0, S1, S2, S3,... , S7 each represents one view, T0, T1, T2, T3,... , T100 represents the time direction.

Referring to FIG. 1, it can be seen that hierarchical B-picture structures defined in H.264 / AVC are used in each view to perform prediction encoding in the time direction. Then, the images of the first time T0 of each view are separated from each other by 8 frames in the time direction, that is, the images T8, T16, T24,. The image at is only performing prediction (inter-view prediction) from neighboring views. For example, time T0, T8, T16, T24,... In S2 view from S0 view, S1 view from S0 and S2 view, S4 view from S2 view, S3 view from S2 and S4 view, S6 view from S4 view, S5 view from S4 view and S6 From the view, the S7 view performs the prediction from the S6 view because it is the last view. Also, in every second view (i.e., S1, S3, S5, S7), inter-view prediction is performed along with intra-view prediction in the time direction at other times (T1, T2, T3, ..., T7, T9, T10, ...). do. That is, in addition to the intra-view prediction in the temporal direction, the S1 view performs the inter-view prediction from the S0 and S2 views, the S3 view from the S2 and S4 views, and the S5 view from the S4 and S6 views.

In multi-view video coding, which is currently being standardized, it is proposed to use a method as defined in H.264 / AVC as a method of predicting motion information, for example, a motion vector (MV). The motion vector (MV) prediction method specified in H.264 / AVC is a motion vector (MV _C , Motion Vector) of a current block obtained through motion estimation (ME) in the time direction. of Current block) is based on the assumption that it is similar to the motion vector of the neighboring block.

More specifically, in H.264 / AVC, the motion vector prediction value MV _P is obtained by using the motion vectors of neighboring blocks MV _N , and then the motion vector of the current block is obtained. MV _C ) is obtained by subtracting the motion vector prediction value MV _P from the motion vector difference value MV _D. The motion vector difference value MV _D is transmitted to the decoder, not the motion vector MV _C of the current block obtained through motion prediction. The equation for obtaining the motion vector difference value MV _D transmitted from the encoder to the decoder can be expressed by Equation 1 below.

In the decoder, the motion vector prediction value MV _P is obtained in the same manner as in the encoder, and the motion vector difference value MV _D transmitted from the encoder is added to the motion vector prediction value MV _P to obtain a motion vector MV of the current block. _C ) The equation for obtaining the motion vector MV _C of the current block in the decoder can be expressed as Equation 2 below.

If there are several neighboring blocks of the current block in H.264 / AVC, the median of all or some of the motion vectors MV _N of the neighboring blocks is obtained, and the motion vector prediction value MV _P of the current block is obtained. It is prescribed to use. In addition, in H.264 / AVC, not only the size of the neighboring block is 16x16 pixels, which are macroblocks, but also 16x8 pixels, 8x16 pixels, 8x8 pixels, 8x4 pixels, and 4x. The case where it is variable to 8 pixels, 4x4 pixels, etc., and the case where the size of the current block is 16x8 pixel or 8x16 pixel, is also prescribed, respectively. In this case, the specific method of obtaining the motion vector prediction value MV _P in each case is quite complicated and diverse. This is briefly described below.

FIG. 2 is a diagram for describing a method of obtaining a motion vector prediction value MV _P using the motion vector MV _N of neighboring blocks in H.264 / AVC. Referring to FIG. 2, neighboring blocks used to obtain the motion vector prediction value MV _P of the current block are, in principle, blocks 'A', 'B', and 'C'. In this case, the motion vector prediction value MV _P becomes an intermediate value of the motion vectors MV of the blocks 'A', 'B', and 'C', as shown in Equation (3).

Here, MV (A) is a motion vector (MV) of block 'A', MV (B) is a motion vector (MV) of block 'B', MV (C) is a motion vector (MV) of block 'C', MV _P is a motion vector prediction value of the current block, and Median (x, y, z) represents an intermediate value of x, y, and z.

And the method of obtaining the median (Median) is represented by the mathematical equation (4).

Here, Min (x, y) represents a smaller or equal value among x and y, and Max (x, y) represents a greater or equal value among x and y.

FIG. 3 is a diagram for explaining another method of obtaining a motion vector prediction value MV _P using the motion vector MV _N of the neighboring block in H.264 / AVC, wherein the neighboring block having the macroblock size has various shapes. This is the case when it is divided into partition blocks. Referring to FIG. 3, in the case of the neighboring block adjacent to the left side of the current block, the partition block at the top right side within the neighboring block is selected as the block 'A', and in the case of the neighboring block adjacent to the current block upward, Select the block 'B' in the bottom left corner of the neighboring block as the block 'B', and in the case of the neighboring block adjacent to the upper right side of the current block, select the partition block in the bottom left side in the neighboring block as the block 'C' The value obtained by the median operation on the motion vectors of the selected partition blocks 'A', 'B', and 'C' is used as the motion vector prediction value MV _P.

FIG. 4 is a view for explaining another method of obtaining a motion vector prediction value MV _P using the motion vector MV _N of neighboring blocks in H.264 / AVC, as described above with reference to FIGS. 2 and 3. In the example, the motion vector of the block 'C' is not available. Referring to FIG. 4, when the motion vector of the block 'C' is not available, the motion vector prediction value MV _P is obtained using the motion vector of the block 'D' instead of the block 'C'. That is, in the example of FIG. 4, the value obtained by the median operation on the motion vectors MV of the blocks 'A', 'B', and 'D' becomes the motion vector prediction value MV _P.

And when the motion vector MV _N of each of the neighboring blocks (blocks 'A', 'B', 'C', and / or 'D') described above is not available, or the neighboring blocks are in the intra prediction mode. In case of a block or a block which does not use the prediction in the time direction, the motion vector (MV) of the neighboring block is set to (0.0, 0.0) and the reference image number is set to '-1' so that the motion vector prediction value ( MV _P )

In addition, as illustrated in FIG. 5A, when the current block is in 8 × 16 pixel mode, the motion vector prediction value MV _P of the block 'A' is used as the motion vector prediction value MV _P of the 8 × 16 block located on the left side. Only using the motion vector prediction value MV _P of the 8 × 16 block located on the right side, using only the motion vector MV of the block 'C'. As shown in FIG. 5B, when the current block is in the 16 × 8 pixel mode, the motion of the block 'B' is used as the motion vector prediction value MV _P of the block having the size of 16 × 8 pixels located above. The vector MV is used, and the motion vector MV of the block 'A' is used as the motion vector prediction value MV _P of a block having a size of 16 × 8 pixels located below the macroblock.

Also, in the above example, when the current block is located at the top of the current image or slice, the blocks 'B' and 'C' are located outside the boundary of the image or slice. In this case, since the blocks 'B' and 'C' are not available, the motion vector MV of the block 'A' is used as the motion vector prediction value MV _P of the current block.

In the above example, even if all of the motion vectors of the blocks 'A', 'B', and 'C' are available, the current block among the motion vectors of the blocks 'A', 'B', and 'C' is available. When there is only one block referring to the same image as this reference image, the motion vector of the block is regarded as the motion vector predicted value MV _P.

In addition, in H.264 / AVC, there is a skip macroblock mode for performing motion compensation (MC) without transmitting macroblock information at all. Here, the motion vector prediction value (MV _P ) is obtained and encoded. Perform And H.264 / AVC, to the spatial direct mode (Direct Mode Spatial) is present, here too by using a motion vector prediction value (MV _P) in the neighboring blocks calculate the current block in the motion vector prediction value (MV _P).

However, since multi-view video coding performs prediction in the view direction along with the prediction in the time direction, two types of motion vectors can be distinguished according to the direction of the motion prediction. That is, the motion vector when the motion prediction is performed in the same view and obtained by referring to images of different time zones and the motion vector is obtained by referring to the images of other views by performing motion prediction in the view direction at the same time. There is. As in the latter case, a motion vector obtained by performing motion prediction in the view direction is also called a disparity vector (DV). Only one of these two kinds of motion vectors may exist in one image, or two kinds of motion vectors may exist simultaneously.

In multi-view video coding, which is currently under international standardization, when an image (picture or slice) includes both kinds of motion vectors, for example, the motion vector MV _C of the current block and the motion vector of the neighboring block ( Even when the directions of images referenced by MV _N ) are different, motion vector prediction values MV _P are obtained in the same manner as defined in H.264 / AVC. That is, the motion vector prediction value MV _P is obtained using only the motion vectors MV _N of neighboring blocks regardless of the direction of the image referenced by the motion vector. However, when the same method as defined in H.264 / AVC is applied uniformly to the motion vector prediction method in multi-view video coding, the following problem may occur.

In performing the prediction, if the motion vector MV _{N of the} neighboring block is obtained by performing the prediction in the time direction (that is, when the reference image of the neighboring block is a different time image in the same view), When the motion vector MV _C is obtained by performing prediction in the view direction (that is, when the reference picture of the current block is a picture of another view in the same time zone), the motion vector MV _N and the current block of the neighboring block are present. The motion vectors MV _C of the blocks are different from each other. In such a case, if the motion vector prediction value MV _P is obtained using only the motion vectors MV _N of neighboring blocks by applying the method specified in H.264 / AVC, the motion vector MV _C of the current block is obtained. ) Is not an accurate prediction. As a result, the amount of information (entropy) for the motion vector difference value MV _D , which is obtained by differentiating the motion vector prediction value MV _P and the motion vector MV _N of the current block, increases, thereby decreasing the efficiency of predictive coding. .

Similarly, even when the motion vector MV _N of the neighboring block is obtained by performing prediction in the view direction, and the motion vector MV _C of the current block is obtained by performing the prediction in the time direction, the motion vector of the neighboring block. (MV _N ) and the motion vector MV _C of the current block are different from each other. Even in such a case, if the method specified in H.264 / AVC is applied as it is, accurate prediction of a motion vector may not be achieved.

Therefore, in the present invention, the characteristics of the motion vector MV _C of the current block and the motion vector MV _N of the neighboring block (whether the motion prediction is performed in the view direction or the motion prediction in the time direction) are the same. The present invention proposes a prediction method of a motion vector in a multiview video coding and a multiview image encoding / decoding method and apparatus using the same, even when the characteristics are different.

According to an aspect of the present invention, a method of obtaining a motion vector predictor in multiview video coding according to an embodiment of the present invention may correspond to motion vectors of neighboring blocks of a current block and time direction of the current block. The motion vector prediction value is obtained by using one or more motion vectors adaptively selected from the motion vectors of the block and the motion vectors of the block corresponding to the view direction of the current block.

According to an aspect of the embodiment, the motion vector prediction value may be a median value of the selected motion vector or a median value of the selected motion vector and (0, 0). Alternatively, the selected motion vector may be one, and the motion vector prediction value may be the same as the selected motion vector.

According to another aspect of the present invention, there is provided a method of obtaining a motion vector prediction value in a multiview video coding, wherein a reference picture of a current block is an image of another view in the same time zone as the current picture, and the current block The motion vector prediction value is obtained by using one or more motion vectors adaptively selected from the motion vectors of the neighboring block of and the motion vectors of the corresponding block in the time direction of the current block.

According to an aspect of the embodiment, the motion vector prediction value may be obtained using a motion vector of a block belonging to the same view as the view to which the reference image of the current block belongs among the neighboring blocks and the time direction corresponding block. have.

According to another aspect of the embodiment, if there is one among the neighboring blocks that the reference image belongs to the same view as the view belonging to the reference image of the current block, the motion vector prediction value using the motion vector of the corresponding neighboring block Can be obtained.

According to another aspect of the embodiment, if there is no block belonging to the same view as the view belonging to the reference image of the current block among the corresponding block in the time direction, the motion using the motion vector of the neighboring block The vector prediction value can be obtained.

According to another aspect of the present invention, there is provided a method of obtaining a motion vector prediction value in multiview video coding, wherein a reference picture of a current block is an image of a different time zone in the same view as the current picture. The motion vector prediction value is obtained by using one or more motion vectors adaptively selected from motion vectors of neighboring blocks of the block and motion vectors of the block corresponding to the view direction of the current block.

According to an aspect of the embodiment, the motion vector prediction value may be obtained using a motion vector of a block belonging to the same time zone that the reference picture belongs to the reference picture of the current block among the neighboring block and the view direction corresponding block. have.

According to another aspect of the embodiment, if there is one among the neighboring blocks, the reference image belongs to the same time zone as the time zone to which the reference image of the current block belongs, the motion vector prediction value using the motion vector of the corresponding neighboring block Can be obtained.

According to another aspect of the embodiment, if there is no block in the view direction corresponding block that the reference image belongs to the same time zone as the time zone to which the reference image of the current block belongs, the motion using the motion vector of the neighboring block A vector prediction can be obtained.

According to another aspect of the present invention, a method of obtaining a motion vector prediction value in multi-view video coding includes: motion vectors of neighboring blocks of a current block; and motion vectors of a time direction corresponding block of the current block. Among the motion vectors of the block corresponding to the view direction of the current block, whether the reference image of the current block is an image of another view in the same time zone as the current image or an image of another time zone in the same view as the current image. Accordingly, the motion vector prediction value is obtained using the adaptively selected motion vector.

According to another aspect of the present invention, there is provided a decoding method of a multiview image, the method comprising: obtaining a motion vector prediction value using the motion vector prediction method according to the embodiment of the present invention; and And calculating a motion vector by adding a vector prediction value and a motion vector difference value of the input current block.

According to another aspect of the present invention, a method of decoding a multiview image according to another embodiment of the present invention determines whether to use a motion vector of a time direction corresponding block and a motion vector of a view direction corresponding block when obtaining a motion vector prediction value. Based on the information indicating, one or more motion vectors adaptively selected from the motion vectors of the neighboring blocks of the current block, the motion vectors of the time-direction corresponding blocks of the current block, and the motion vectors of the view-direction corresponding blocks of the current block are selected. Obtaining the motion vector prediction value by using the motion vector prediction value, and calculating a motion vector of the current block by summing the motion vector prediction value and the input motion vector difference value of the current block.

According to an aspect of the embodiment, the information may be information included in a sequence header, information included in a slice header, and / or encoding information about a block. In this case, the information may be a variable length code, a fixed length code, or an Exp-Golomb code.

According to another aspect of the embodiment, in the step of obtaining the motion vector prediction value, when the information is information indicating that the motion vector of the corresponding time direction block and the motion vector of the corresponding view direction block are not used, If only the motion vector is used and the information indicates that the motion vector of the corresponding time direction block and the motion vector of the corresponding view direction block are used, the motion vector of the neighboring block and the motion vector of the time direction corresponding block are used. And one or more motion vectors adaptively selected from the motion vectors of the view direction corresponding blocks.

According to another aspect of the present invention, there is provided a method of decoding a multiview image, based on information indicating whether to use a motion vector of a time direction corresponding block when obtaining a motion vector prediction value. Obtaining a motion vector prediction value of the current block by using a motion vector of at least one block adaptively selected from the motion vector of the neighboring block of the block and the motion vector of the corresponding block in the time direction of the current block, and the motion vector prediction value; Calculating a motion vector of the current block by summing input motion vector difference values of the current block.

According to an aspect of the embodiment, in the step of obtaining the motion vector prediction value, when the information indicates that the motion vector of the corresponding block in the time direction is not used, only the motion vector of the neighboring block is used and the information is used. If is information indicating that the motion vector of the corresponding time direction block is used, the motion vector of at least one block adaptively selected from the motion vector of the neighboring block and the motion vector of the time direction corresponding block may be used.

According to another aspect of the present invention, there is provided a decoding method of a multiview image, based on information indicating whether to use a motion vector of a view direction corresponding block when a motion vector prediction value is obtained. Obtaining a motion vector prediction value of the current block by using one or more motion vectors adaptively selected from a motion vector of a neighboring block of the motion block and a motion vector of the corresponding block in the view direction of the current block, and inputting the motion vector prediction value to the motion vector prediction value Calculating a motion vector of the current block by summing motion vector difference values of the current block.

According to an aspect of the embodiment, in the step of obtaining the motion vector prediction value, when the information is information indicating that the motion vector of the corresponding view direction block is not used, only the motion vector of the neighboring block is used. When the information indicates that the motion vector of the corresponding view direction block is used, one or more motion vectors adaptively selected from the motion vector of the neighboring block and the motion vector of the view direction corresponding block may be used.

According to another aspect of the present invention, there is provided a method of encoding a multiview image, the method comprising: obtaining a motion vector prediction value using the motion vector prediction method according to the embodiment; and the motion vector prediction value and the Subtracting a motion vector of the current block to obtain a motion vector difference value of the current block.

According to another aspect of the present invention, a method of encoding a multiview image according to another embodiment of the present invention determines whether to use a motion vector of a time direction corresponding block and a motion vector of a view direction corresponding block when obtaining a motion vector prediction value. Generating information indicating, based on the information, an adaptively selected from among a motion vector of a neighboring block of the current block, a motion vector of a time direction corresponding block of the current block, and a motion vector of a view direction corresponding block of the current block; Obtaining a motion vector prediction value using at least one motion vector, and subtracting the motion vector prediction value and the motion vector of the current block to obtain a motion vector difference value of the current block.

According to another aspect of the present invention, there is provided a method of encoding a multiview image, generating information indicating whether to use a motion vector of a block corresponding to a time direction when obtaining a motion vector prediction value, Obtaining a motion vector prediction value using at least one motion vector adaptively selected from a motion vector of a neighboring block of the current block and a motion vector of a time-direction corresponding block of the current block, based on the information, and the motion vector prediction value And subtracting the motion vector of the current block to obtain a motion vector difference value of the current block.

According to another aspect of the present invention, there is provided a method of encoding a multiview image, generating information indicating whether to use a motion vector of a view direction corresponding block when a motion vector prediction value is obtained; Obtaining a motion vector prediction value using at least one motion vector adaptively selected from a motion vector of a neighboring block of the current block and a motion vector of a view direction corresponding block of the current block based on the information, and the motion vector prediction value And subtracting the motion vector of the current block to obtain a motion vector difference value of the current block.

The syntax of multi-view video coding according to an embodiment of the present invention for achieving the above technical problem is to use a motion vector of a time-direction corresponding block and / or a motion vector of a view-direction corresponding block when obtaining a motion vector prediction value. Contains information indicating whether or not.

According to an aspect of the embodiment, the syntax may be syntax for a sequence, syntax for a slice, and / or syntax for a macroblock.

The apparatus for decoding a multiview image according to an embodiment of the present invention for achieving the above technical problem is a motion vector predictor for obtaining a motion vector prediction value using the fms motion vector prediction method according to the embodiment of the present invention described above. And an adder for generating a motion vector of the current block by summing the motion vector predicted value and the input motion vector difference value.

According to another aspect of the present invention, an apparatus for decoding a multiview image according to another embodiment of the present invention determines whether to use a motion vector of a time direction corresponding block and a motion vector of a view direction corresponding block when obtaining a motion vector prediction value. On the basis of the input information indicating, the at least one block adaptively selected from the motion vector of the neighboring block of the current block, the motion vector of the time-direction corresponding block of the current block and the motion vector of the view-direction corresponding block of the current block A motion vector predictor for obtaining the motion vector predictor using a motion vector, and an adder for calculating the motion vector of the current block by summing the motion vector predictor and the input motion vector difference.

According to another aspect of the present invention, there is provided a decoding apparatus of a multiview image, based on input information indicating whether to use a motion vector of a block corresponding to a time direction when obtaining a motion vector prediction value. A motion vector predictor for obtaining the motion vector prediction value using a motion vector of at least one block adaptively selected from a motion vector of a neighboring block of the current block and a motion vector of the time-direction corresponding block of the current block, and And an adder for calculating a motion vector of the current block by summing a motion vector prediction value and an input motion vector difference value.

According to another aspect of the present invention, there is provided a decoding apparatus for a multiview image, based on input information indicating whether to use a motion vector of a view direction corresponding block when obtaining a motion vector prediction value. A motion vector predictor for obtaining the motion vector prediction value using a motion vector of at least one block adaptively selected from a motion vector of a neighboring block of a current block and a motion vector of the view direction corresponding block of the current block, and the motion And an adder for calculating a motion vector of the current block by adding a vector predicted value and an input motion vector difference value.

According to an aspect of the present invention, there is provided an apparatus for encoding a multiview image, including: a motion vector predictor for obtaining a motion vector prediction value using the motion vector prediction method according to the embodiment of the present invention; A motion predictor for obtaining a motion vector of the current block using a time direction reference image including the time direction corresponding block and a view direction reference image including the view direction corresponding block, and the motion vector and the motion vector prediction value And an adder for calculating a motion vector difference value of the current block by subtracting.

According to another aspect of the present invention, an apparatus for encoding a multiview image according to another embodiment of the present invention determines whether to use a motion vector of a time direction corresponding block and a motion vector of a view direction corresponding block when obtaining a motion vector prediction value. A usage determination unit for generating information indicating, based on the information, a motion vector of a neighboring block of the current block, a motion vector of a time direction corresponding block of the current block, and a motion vector of a view direction corresponding block of the current block A motion vector predictor for obtaining the motion vector predictor using an adaptively selected motion vector of the at least one block; and subtracting the motion vector predictor and the motion vector of the current block to subtract the motion vector difference of the current block. Include an adder to find.

According to another aspect of the present invention, an apparatus for encoding a multiview image according to another embodiment of the present invention is used to generate information indicating whether to use a motion vector of a time-direction corresponding block when obtaining a motion vector prediction value. A determination unit, based on the information, a motion for obtaining the motion vector prediction value using one or more motion vectors adaptively selected from the motion vectors of the neighboring blocks of the current block and the motion vectors of the corresponding blocks in the time direction of the current block; And a vector predictor and an adder for subtracting the motion vector predicted value and the motion vector of the current block to obtain a motion vector difference value of the current block.

According to another aspect of the present invention, an apparatus for encoding a multiview image according to another embodiment of the present invention is used to generate information indicating whether to use a motion vector of a view direction corresponding block when a motion vector prediction value is obtained. A determination unit, based on the information, a motion for obtaining the motion vector prediction value using one or more motion vectors adaptively selected from the motion vectors of the neighboring blocks of the current block and the motion vectors of the corresponding block in the view direction of the current block; And a vector predictor and an adder for subtracting the motion vector predicted value and the motion vector of the current block to obtain a motion vector difference value of the current block.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the present invention. Since the embodiments described below are for the purpose of illustrating the technical idea of the present invention, the technical idea of the present invention should not be construed as being limited by the embodiments. In the description of the following embodiments, the names of each component may be referred to as other names in the art, and if they have functional similarity and identity, they may be regarded as equivalent to the embodiments of the present invention even though other names are used. have. Similarly, even if an embodiment in which the configuration on the drawings is partially modified is adopted, both can be regarded as an equivalent configuration if there is functional similarity and identity. Reference numerals added to the respective components in the description of the embodiment and the drawings are merely described for convenience of description.

In multi-view video coding according to the present invention, when obtaining the motion vector prediction value MV _P of the current block, one or more of the following three motion vectors MV are used.

(1) A motion vector (MV _N ) of neighboring blocks of the current block.

(2) a motion vector MV _T of a corresponding block in a time direction, that is, a block having the same position as that of the current block in an image at another time (hereinafter, referred to as a 'time direction reference image') in the same view as the current image and Vector of motion of surrounding blocks.

(3) the motion vector MV _V of the corresponding view direction block, that is, a block having the same position as that of the current block in an image of another view at the same time as the current image (hereinafter, referred to as a 'view direction reference image') and Motion vector (MV) of neighboring blocks.

In the case of an image to be encoded as an inter-picture, the prediction coding efficiency can be improved if an accurate motion vector prediction value MV _P can be obtained. That is, when the accurate motion vector prediction value MV _P is obtained, the amount of information of the motion vector difference value MV _D transmitted to the decoder can be reduced by that much, so that the compression ratio of the image can be increased. As described above, in multi-view video coding, there are two cases in which a motion vector is obtained by performing prediction in the temporal direction according to its characteristics and a case where the motion vector is obtained by performing prediction in the view direction. This means on the one hand that more information about the current block can be used for predictive encoding. If a lot of information is used for prediction encoding, a more accurate motion vector prediction value MV _P can be obtained, and the amount of information of the motion vector differential value MV _D can be reduced by that much.

In the motion vector prediction method in multi-view video coding according to the present invention, a motion vector prediction value (MV _P ) is obtained by selecting some of the three types of motion vectors (MV _N , MV _T , MV _V ) described above. Preferably, the motion vector having the same characteristics as that of the motion vector MV _C of the current block is selected. As described above, since the block that can be used to obtain the motion vector prediction value MV _P is not limited to the neighboring block, the characteristics of the motion vector MV _C of the current block and the motion vector MV _N of the neighboring block are the same. In addition to the case, even when the characteristic is different, a more accurate motion vector prediction value MV _P can be obtained. By obtaining a more accurate motion vector prediction value MV _P , the amount of information of the motion vector difference value MV _D can be reduced and the coding efficiency can be improved. This will be described below with more specific example.

6A, 6B and 6C are all defined in H.264 / AVC with the image " Ballroom " which is one of the test image sequences in the standardization process of multi-view video coding, with the size of the image group in the time direction being 12; This is an example of the case of encoding using a conventional motion vector prediction method. 6A shows a part of the overall prediction structure, in which a type and number of each picture and a reference relationship between the pictures are shown. FIG. 6B is an enlarged view of motion information of some macroblocks of the image B ₁ (1) corresponding to the time T0 of the S1 view in the prediction structure shown in FIG. 6A, and FIG. 6C is the prediction shown in FIG. 6A. In the structure, the motion information of some macroblocks of the image b ₄ (21) corresponding to the T1 time of the S1 view is enlarged and displayed. In FIG. 6B and FIG. 6C, the information in each macroblock is sequentially obtained from the motion vector (MV), the motion vector prediction value (MV _P ), the motion vector difference value (MV _D, ), and the reference image (ie, reference). Number of the image).

First, referring to FIG. 6B, since the macroblocks 12 and 14 of the image B ₁ (1) have a reference image number of 0, the motion prediction refers to the image I ₀ (0) corresponding to the time T0 of the S0 view. The motion vector is (17.8, 2.0). Since the macroblocks located above and to the right of the macroblocks 12 and 14, that is, the macroblocks 12 and 13 and 13 and 13, information on the motion vector MV does not exist (not shown in the drawing). Denotes N / A), the motion vector prediction value MV _P is a macroblock located to the left of the macroblocks 12 and 14, that is, a motion vector MV of the macroblocks 11 and 14 ( 11.8, 2.5). Therefore, the motion vector difference value MV _D = MV-MV _P for the macroblocks 12 and 14 of the image B ₁ (1) is (6.0, -0.5).

6C, since the macroblocks 12 and 14 of the image b ₄ 21 have a reference image number 8, the motion prediction refers to the image B ₃ (8) corresponding to the time T1 of the S0 view. The motion vector is (18.3, 2.0). The motion vectors MV of the macroblocks 11 and 14 positioned on the left side of the macroblocks 12 and 14 and the macroblocks 12 and 13 positioned on the upper side thereof are both (0.0, 0.0) and are located on the upper right side. The macroblocks 13 and 13 do not have a motion vector MV. In this case, as described with reference to FIG. 4, the motion vector prediction value MV _P is obtained using the motion vector MV of the macroblocks 11 and 13 positioned at the block 'D', that is, at the upper left. . Since the motion vectors MV of the macroblocks 11 and 13 are (0.0, -0.3), the middle of the motion vectors MV of the macroblocks 11, 14, 12, 13, and 11, 13 is the same. The value (0.0, 0.0) becomes the motion vector prediction value MV _P. Therefore, the motion vector difference value MV _D for the macroblocks 12 and 14 of the image b ₄ (21) is (18.3, 2.0).

6B and 6C, the motion vector difference values MV _D of the current blocks 12 and 14 are (6.0, -0.5) and (18.3, 2.0), respectively, which are much larger. There may be various causes of this, but one of them is the neighboring blocks 11 and 14 used to obtain the reference image and the motion vector prediction value MV _P of the current blocks 12 and 14 in the case of FIG. 6B. Although the reference image of is the same as the image I ₀ (0) of the time T0 of the view S0, in the case of FIG. 6C, the reference image of the current blocks 12 and 14 is the image B ₃ (8) of the time T1 of the view S0, but the movement The reference images of the neighboring blocks (11, 13), (12, 13), and (11, 14) used to obtain the vector prediction value MV _P are the image B ₁ (1) of the time T0 of the S1 view, The characteristics of the reference image of the block and the reference image of the neighboring block are different from each other.

Therefore, as in the case of FIG. 6C, when the reference image of the current block and the reference image of the block used when obtaining the motion vector prediction value MV _P are different from each other, the motion vector difference value MV _D having a relatively large value is obtained. ) Is encoded and transmitted, resulting in inferior encoding efficiency. However, in the motion vector prediction method in multi-view video coding according to the present invention, any one or more of the motion vectors MV _S , MV _T and / or MV _V described above (1), (2), and (3). This problem can be solved because the motion vector prediction value MV _P is obtained by selecting " This will be described in detail with reference to FIGS. 6A to 6C.

Assume that the current block to be encoded is macroblocks 12 and 14 of FIG. 6C. In this case, the reference image of the current blocks 12 and 14 is the image B ₃ (8) located in the S0 view at the same time T1 as the time of the current image b ₄ (21). That is, in motion prediction, the current blocks 12 and 14 refer to the image in the view direction (S0 view). Also, in FIG. 6B, macroblocks 12 and 13 positioned at the same position as the current block also refer to image I ₀ (0) positioned in the S0 view at the same time T0. As a result, in the motion prediction, the direction of the image referred to by the macroblocks 12 and 14 of the current image b ₄ (21) and the image referred to by the macroblocks 12 and 14 of the image B ₁ (1) both refer to the same view. As the image, the characteristics of the reference image are the same.

In such a case, coding efficiency can be improved by obtaining a motion vector prediction value using the motion vector MV _T of the corresponding time direction block in the same view. In the above example, when the motion vector MV (17.8, 2.0) of the macroblocks 12 and 14 of the image B ₁ (1) is used as the motion vector prediction value MV _P , the motion vector difference value MV _D ) Becomes (0.5, 0.0). These results show that the prediction coding efficiency can be improved by reducing the amount of information on the motion vector difference value MV _D as compared with the conventional method in which the motion vector difference value MV _D is (18.3, 2.0).

After all, the method of predicting a motion vector in multi-view video coding according to the present invention includes a current block (the meaning of the block is not limited to a macroblock of 16 × 16 units and may be a block of various sizes, or a predetermined size). The prediction coding efficiency is improved by using a prediction method different from the conventional method of obtaining the motion vector prediction value MV _P using only the motion vector MV _N of neighboring blocks of a region having a size (which may be a region). You can. That is, in the method of predicting a motion vector in multi-view video coding according to the present invention, (1) the motion vector (MV _N ) of the neighboring blocks of the current block, and (2) the 'time of the temporal reference image' for the current image. Selects a motion vector from the motion vector MV _T of the 'direction corresponding blocks' and / or (3) the motion vector MV _V of the 'view direction corresponding blocks' of the 'view direction reference image' for the current image, The motion vector prediction value MV _P is obtained using the motion vector.

Motion vector prediction (MV _P Reference blocks that can be used to find

FIG. 7 illustrates reference blocks that can be used to obtain a motion vector prediction value MV _P of a current block according to the present invention. The reference block may be, for example, a macroblock (MB). Alternatively, as another example, an 8 × 8 pixel block or a 4 × 4 pixel block may be used, and a block having an arbitrary size used in H.264 / AVC (16 × 8, 8 × 16, 8 × 4, 4 ×). 8) or the like.

First, FIG. 7 will be described in more detail with reference to a prediction structure in multi-view video coding shown in FIG. 1. For example, if the current image (image including reference number 10 of FIG. 7) is image b ₄ at time T1 of view S1 in FIG. 1, the image including reference number 12 is time T0 of view S1 in FIG. 1. image is the image B ₃ at T2 time of S1 view in Figure 1 including the image B _1, reference number 14 and in this. The image including reference number 16 becomes image B ₃ at time T1 of view S0 in FIG. 1, and the image including reference number 18 becomes image B ₃ at time T1 of view S2 in FIG. 1.

If the current image is the image B ₃ at the time T2 of the S1 view in FIG. 1, the image including the reference number 12 becomes the image B ₁ at the time T0 of the S1 view in FIG. 1, and the image including the reference number 14. Becomes image B ₂ at time T4 of the S1 view in FIG. 1. The image including reference number 16 becomes image B ₂ at time T2 of the S0 view in FIG. 1, and the image including reference number 18 becomes image B ₂ at T2 time of the S2 view in FIG. 1.

Referring to FIG. 7, reference numeral 10 denotes neighboring blocks of a current block used when obtaining a motion vector prediction value MV _P in the existing H.264 / AVC. According to an embodiment of the present invention, the motion vector prediction value MV _P may be obtained using the motion vectors MV _N of the neighboring blocks 'A', 'B', 'C', and 'D'.

Reference numerals 12 and 14 are 'time direction corresponding blocks' of the 'time direction reference image'. More specifically, reference numeral 12 denotes a block of a reference image of time direction list 0 (list0) or a block of a reference image temporally preceding in the same view as the current image, and reference numeral 14 denotes a time direction list in the same view as the current image. A block of the reference picture of 1 (list1) or a block of the reference picture that is backward in time. According to an embodiment of the present invention, blocks 'TF0', 'TF1', 'TF2', 'TF3', which are 'time direction corresponding blocks' in a 'time direction reference image' that is temporally ahead of the current image in the same view, 'TF4', 'TF5', 'TF6', 'TF7', and 'TF8' and the 'time direction reference block' in the temporal reference image that is later than the current image. A motion vector prediction value MV _P can be obtained using the motion vectors MV _T of 'TB2', 'TB3', 'TB4', 'TB5', 'TB6', 'TB7', and 'TB8'. In reference numerals 12 and 14, blocks 'TF0' and 'TB0' are blocks at the same position as the current block, respectively, and the remaining blocks are peripheral blocks thereof.

Also, reference numeral 16 and reference numeral 18 denote 'view direction corresponding blocks' of the 'view direction reference image'. More specifically, reference numeral 16 denotes a block of the reference image in the view direction list 0 (list0) or a block of the reference image in the front view at the same time as the current image, and reference numeral 18 denotes the view direction at the same time as the current image. A block of the reference picture in list 1 or a block of the reference picture in the rear view is shown. According to an embodiment of the present invention, blocks' VF0 ',' VF1 ',' VF2 ',' which are 'view direction corresponding blocks' in a' view direction reference image 'in a view ahead of the view to which the current image belongs at the same time. VF3 ',' VF4 ',' VF5 ',' VF6 ',' VF7 'and' VF8 'and the block' view direction correspondence block 'in the' view direction reference image 'in the rear view than the view to which the current image belongs Motion vector prediction value (MV _V ) using motion vectors (MV _V ) of VB0 ',' VB1 ',' VB2 ',' VB3 ',' VB4 ',' VB5 ',' VB6 ',' VB7 ', and' VB8 '. _P ) can be obtained. In reference numerals 16 and 18, blocks 'VF0' and 'VB0' are blocks at the same position as the current block, and the remaining blocks are peripheral blocks thereof.

In obtaining the motion vector prediction value MV _P of the current block, not all of the above-mentioned reference blocks or motion vectors of the reference block that can be used to obtain the motion vector prediction value MV _P are described. It is desirable to make appropriate selections, using only the motion vectors of the selected reference blocks. That is, the reference block that is actually used in obtaining the motion vector prediction value MV _P , that is, the set of the selected reference blocks includes all the reference blocks shown in FIG. 7 in one set S = {'A', 'B', 'C', 'D', 'TF0', 'TF1', 'TF2', 'TF3', 'TF4', 'TF5', 'TF6', 'TF7', 'TF8', 'TB0', 'TB1 ',' TB2 ',' TB3 ',' TB4 ',' TB5 ',' TB6 ',' TB7 ',' TB8 ',' VF0 ',' VF1 ',' VF2 ',' VF3 ',' VF4 ', 'VF5', 'VF6', 'VF7', 'VF8', 'VB0', 'VB1', 'VB2', 'VB3', 'VB4', 'VB5', 'VB6', 'VB7', 'VB8''}, Only reference blocks belonging to any one subset of all subsets of S can be used. For example, by selecting a set S ₁ = {'A', 'B', 'C', 'D', 'VF0', 'VB0', 'TF0', 'TB0'}, which is one of a subset of S, A motion vector prediction value MV _P may be obtained using all or part of the reference blocks included in the set S ₁ .

Preferably, based on the characteristics of the image referenced by the motion vector MV _C of the current block in motion prediction, a subset of the set S, which is a set of reference blocks that are actually used to obtain a motion vector prediction value MV _P , is obtained. It can be limited. For example, (A) the motion vector (MV _C ) of the current block refers to the image in the view direction and (B) the motion vector (MV _C ) of the current block refers to the image in the time direction The subset of the set S to be selected may be more specifically defined.

8 is a diagram illustrating another example of a prediction structure in multiview video coding, which is a case of three consecutive time frames in three views. Hereinafter, the "time direction reference image" and the "view direction reference image" will be described in more detail with reference to FIG. 8.

Referring to FIG. 8, when the image b ₄ 21 at the time T1 of the S1 view is the current image, the reference image of the time direction list 0 (list0) of the current image b ₄ 21, or the temporally advanced reference image is The reference image B ₁ at time T0 of the S1 view, and the reference image or temporally backward reference image of the temporal list 1 of the current image b ₄ 21 is image B ₃ at time T2 of the S1 view. The reference image of the view direction list 0 (list0) of the current image b ₄ (21) or the view that is earlier than the view to which the current image belongs is the image B ₃ at the time T1 of the S0 view, and the current image b ₄ ( The reference image of the reference image of the view direction list 0 (list0) of 21) or the view behind the view to which the current image belongs is image B ₃ at the time T1 of the S2 view.

9 is a view illustrating the drawings of FIGS. 7 and 8 together. Hereinafter, with reference to FIG. 9, it demonstrates concretely about limiting the subset of the said set S selected by the case of (a) and (b) more specifically.

(A) When the current block refers to the 'view direction reference image' in the motion prediction, that is, when the motion vector (MV _C ) of the current block refers to the view direction (in the example described below, in case of (i) ii), including the case of (iii))

(i) The subset S ₂ of the set S selected in this case is the 'time direction corresponding block' in the 'time direction reference image' together with the neighboring blocks of the current block (i.e. blocks of reference numerals 12 and 14 in FIG. 7). It may be limited to be configured as). And the motion vector prediction value MV _P is calculated using the motion vectors MV _N and / or MV _T of all or part of the block belonging to this set S ₂ .

For example, in FIG. 9, if the current image is image b ₄ at time T1 of view S1 and the current block refers to image B ₃ at time T1 of view S0, set S ₂ is a block that is a neighboring block of the current block. 'a', 'b', 'C', and with a 'D', the image of the T2 time of the image b ₁ and S1 view of the T0 time of S1 view of the current time reference direction image of image b ₄ b ₃ Blocks 'TF0', 'TF1', 'TF2', 'TF3', 'TF4', 'TF5', 'TF6', 'TF7', 'TF8', 'TB0', It consists of 'TB1', 'TB2', 'TB3', 'TB4', 'TB5', 'TB6', 'TB7', and 'TB8'. The motion vector prediction value MV _P may be obtained using the motion vectors MV _N and / or MV _T of all or part of the blocks constituting the set S ₂ .

(ii) According to the embodiment, the set S ₂ used when obtaining the motion vector prediction value MV _P may be more specifically limited to the set S ₃ described later, which is a subset thereof. For example, the set S ₃ may be limited to consist of blocks having a motion vector referenced in the same view direction as the view referenced by the motion vector MV _C of the current block. The motion vector prediction value MV _P is obtained using the motion vectors MV _N and / or MV _T of all or some of the blocks belonging to the set S ₃ . For example, in FIG. 9, if the motion vector MV _C of the current block refers to the S0 view, the set S ₃ such that the motion vector is composed only of blocks referring to the S0 view among the blocks constituting the set S _2. The motion vector prediction value MV _P may be obtained using the motion vectors MV _N and / or MV _T of all or some of the blocks belonging to the set S ₃ .

(iii) And when there is no block belonging to the set S ₃ among the blocks belonging to the set S ₂ , the motion vector prediction value MV _P is obtained using only the motion vector MV _N of the neighboring block, as in the conventional method. You can get it. For example, if the motion vector MV _C of the current block refers to the view S0, and none of the blocks constituting the set S ₂ has the motion vector MV _N and / or MV _T referring to the view S0. The motion vector prediction value MV _P may be obtained using the motion vectors MV _N of all or some of the neighboring blocks 'A', 'B', 'C', and 'D'.

(B) When the current block refers to the 'time direction reference image' in motion prediction, that is, when the motion vector (MV _C ) of the current block refers to the time direction (in the example described below, in case of (i) ii), including the case of (iii)).

(i) The subset S ₄ of the set S selected in this case together with the neighboring blocks of the current block is the 'view direction corresponding block' in the 'view direction reference image' (ie blocks of reference numerals 16 and 18 in FIG. 7). It may be limited to be configured as). The motion vector prediction value MV _P is obtained using the motion vectors MV _N and / or MV _V of all or part of the block belonging to the set S ₄ .

For example, in FIG. 9, if the current picture is picture b ₄ at time T1 of view S1 and the current block refers to picture B ₁ at time T0 of view S1, set S ₄ is a block that is a neighboring block of the current block. 'a', 'b', 'C', and with a 'D', the image of the T1 time of the picture b ₃ and view S2 in the T1 time of the S0 view the current image b, see the view direction of the _fourth image b ₃ Blocks 'VF0', 'VF1', 'VF2', 'VF3', 'VF4', 'VF5', 'VF6', 'VF7', 'VF8', 'VB0', It consists of 'VB1', 'VB2', 'VB3', 'VB4', 'VB5', 'VB6', 'VB7', and 'VB8'. The motion vector prediction value MV _P may be obtained using the motion vectors MV _N and / or MV _V of all or part of the blocks constituting the set S ₄ .

(ii) According to the embodiment, the set S ₄ used when obtaining the motion vector prediction value MV _P may be more specifically limited to the set S ₅ described later, which is a subset thereof. For example, the set S ₅ may be limited to include only blocks having a motion vector referring to an image having a same time zone as the time zone referenced by the motion vector MV _C of the current block. The motion vector prediction value MV _P is obtained using the motion vectors MV _N and / or MV _V of all or part of the block belonging to the set S ₅ . For example, if the motion vector (MV) of the current block refers to the time zone T0 in FIG. 9, the set such that the motion vector (MV) is composed of only the blocks referring to the time zone T0 among the blocks constituting the set S ₄ . may be limited to S _5, and by using all or a portion of the motion vector (MV _N and / or MV _V) of the blocks belonging to the set S ₅ to obtain a motion vector prediction value (MV _P).

(iii) And when there is no block belonging to the set S ₅ among the blocks belonging to the set S ₄ , the motion vector predicted value MV using only the motion vector MV _N of the neighboring blocks of the current block as in the conventional method. _P ) can also be obtained. For example, if the motion vector MV of the current block refers to the time zone T0, and none of the blocks constituting the set S ₄ has the motion vector MV _N and / or MV _V referring to the time zone T0, A motion vector prediction value MV _P may be obtained using the motion vectors MV _N of all or part of blocks 'A', 'B', 'C', and 'D', which are neighboring blocks of the current block.

Motion vector prediction (MV _P Algorithm of the block selection method used to find

According to the present invention, in obtaining the motion vector prediction value MV _P of the current block, the motion vector MV of all or some blocks belonging to one of the subsets of the set S is selected and the selected motion vector is used. As described above, the selected subset of the set S includes, as described above, whether the motion vector MV _C of the current block refers to an image in a view direction (in case of (A)) or an image in a time direction ( It may be more specifically defined based on the case of (b). As described in detail above, the subset of the set S selected in the case of (a) may consist of all or some of the neighboring blocks and / or the time direction corresponding blocks of the current block, and in the case of (b) The subset of the set S selected by may be composed of all or some of the neighboring blocks and / or view direction corresponding blocks of the current block. And a subset of the set S selected according to which (A)-(ii), (A)-(iii), (B)-(ii), and (B)-(iii) corresponds to It may be more limited. Hereinafter, the algorithm applicable to the cases (a)-(ii) and (a)-(iii) mentioned above and the cases (b)-(ii) and (b)-(iii) An example is demonstrated.

FIG. 10 is a flowchart illustrating an example of a method of obtaining a motion vector prediction value MV _P when the motion vector MV _C of the current block refers to an image in a view direction. As described above, in this case the set S ₂ = {'A', 'B', 'C', 'D', 'TF0', 'TF1', 'TF2', 'TF3', 'TF4', ' TF5 ',' TF6 ',' TF7 ',' TF8 ',' TB0 ',' TB1 ',' TB2 ',' TB3 ',' TB4 ',' TB5 ',' TB6 ',' TB7 ',' TB8 ' } Can be selected to be used to obtain the motion vector prediction value MV _P.

Referring to FIG. 10, among the blocks constituting the set S ₂ , an image in a view direction is referred to among motion vectors MV _N of blocks 'A', 'B', 'C', and 'D', which are neighboring blocks. It is determined whether there is any (S11). As a result of the determination, if one or more motion vectors exist, the flow branches to step S14.

In step S12, the blocks 'TF0', 'TF1', 'TF2', 'TF3', 'TF4', 'TF5', 'TF6', 'TF7', 'TF8', 'TB0', Is there any motion vector (MV _T ) of 'TB1', 'TB2', 'TB3', 'TB4', 'TB5', 'TB6', 'TB7', and 'TB8' that refer to the image in the view direction? Determine (S12). As a result of the determination, if one or more motion vectors exist, the flow branches to step S13, and if none exist, branches to step S14.

Subsequently, if at least one of the motion vectors MV _T of the corresponding block in the time direction refers to an image in the view direction, the motion vector prediction value using all or part of the corresponding motion vectors _M V _T is continued. Find (MV _P ). In some embodiments, the motion vector prediction value MV _P may be obtained by using the motion vector MV _N of the neighboring block together with the motion vector MV _T of the corresponding time direction block.

In step S14, that is, at least one of the motion vectors MV _N of the neighboring blocks refers to an image in the view direction, or in the motion vectors MV _T of the time direction corresponding block, the image in the view direction is referred to. If none is found, the motion vector prediction value MV _P of the current block is obtained using all or part of the motion vector MV _N of the neighboring block.

FIG. 11 is a flowchart illustrating an example of a method of obtaining a motion vector prediction value MV _P when the motion vector MV _C of the current block refers to an image in a time direction. As described above, in this case the set S ₄ = {'A', 'B', 'C', 'D', 'VF0', 'VF1', 'VF2', 'VF3', 'VF4', ' VF5 ',' VF6 ',' VF7 ',' VF8 ',' VB0 ',' VB1 ',' VB2 ',' VB3 ',' VB4 ',' VB5 ',' VB6 ',' VB7 ',' VB8 ' } May be selected to be used to obtain the motion vector prediction value MV _P.

Referring to FIG. 11, first, an image in a time direction is referred to among motion vectors MV _N of blocks 'A', 'B', 'C', and 'D', which are neighboring blocks, among blocks constituting the set S ₄ . It is determined whether there is any (S21). As a result of the determination, if at least one of the corresponding motion vectors exists, the flow branches to step S24, and if none exist, the flow branches to step S22.

In operation S22, the blocks 'VF0', 'VF1', 'VF2', 'VF3', 'VF4', 'VF5', 'VF6', 'VF7', 'VF8', 'VB0', Whether any of the motion vectors (MV _V ) of 'VB1', 'VB2', 'VB3', 'VB4', 'VB5', 'VB6', 'VB7', and 'VB8' refers to an image in the time direction It is determined (S22). As a result of the determination, if at least one of the corresponding motion vectors exists, the process branches to step S23, and if none exist, the process branches to step S24.

Subsequently, if at least one of the motion vectors MV _V of the view direction corresponding blocks refers to the image in the time direction, step S23 is used, in which all or part of the corresponding motion vectors MV _V are used. The prediction value MV _P is obtained. In some embodiments, the motion vector prediction value MV _P may be obtained by using the motion vector MV _N of the neighboring block and the motion vector MV _V of the corresponding view direction block.

In step S24, that is, there is at least one reference to an image in a time direction among the motion vectors MV _N of the neighboring blocks, or to refer to an image in the time direction among the motion vectors MV _V of the corresponding view direction blocks. In the absence of any one, the motion vector prediction value MV _P of the current block is obtained using all or part of the motion vector MV _N of the neighboring block.

Motion vector prediction (MV _P Block is used to estimate motion of various magnitudes.

Since multi-view video coding, which is currently being standardized, is basically based on the contents defined in H.264 / AVC, various blocks used to obtain a motion vector prediction value (MV _P ) according to an embodiment of the present invention are also various. Magnitude motion estimation can be used. In this case, the block used when obtaining the motion vector prediction value MV _P , that is, each block belonging to the aforementioned set S has a motion vector MV corresponding to the number of partition blocks. Therefore, when the block used to obtain the motion vector prediction value MV _P uses motion estimation of various sizes, one partition block is selected from the partition blocks of the block, and the motion vector MV of the selected partition block is selected. It is preferable to obtain the motion vector prediction value MV _P using only. In this case, the position of the partition block selected may be appropriately determined in consideration of the prediction encoding efficiency.

Figure 12 shows the case there is a block which is used to obtain the motion vector prediction value (MV _P) divided into multiple partitions, blocks, for obtaining the motion vector prediction value (MV _P) is a partitioning block location is actually used is schematically exemplified. Referring to FIG. 12, the first drawing is a case where a partition block located at the upper right of the block is selected, the second drawing is a case where a partition block located at the lower right of the corresponding block is selected, and the third drawing is a corresponding block. In the case where the partition block located at the lower left of the block is selected, and the fourth figure is a case in which the partition block located at the upper left of the block is selected.

A. When the corresponding block is a block corresponding to reference numeral 10 of FIG. 7, that is, a neighboring block, and the neighboring block is divided into partition blocks of various shapes, the motion vector prediction value MV _P is obtained as follows. Can be.

(i) When the block 'A' of FIG. 7 is divided into partition blocks having various shapes, the motion vector of the partition block located at the uppermost right side of the block 'A' is used.

(ii) When the block 'B' of FIG. 7 is divided into partition blocks having various shapes, the motion vector of the partition block located at the bottom left of the block 'B' is used.

(iii) When the block 'C' of FIG. 7 is divided into partition blocks having various shapes, the motion vector of the partition block located at the bottom left of the block 'C' is used.

(iv) When the block 'D' of FIG. 7 is divided into partition blocks having various shapes, the motion vector of the partition block located at the lower rightmost side of the block 'D' is used.

B. The reference block is a block corresponding to reference numerals 12 and 14 of FIG. 7, that is, a time direction corresponding block, and the time direction corresponding block is divided into partition blocks of various shapes, or the reference block is a reference number of FIG. When the blocks corresponding to 16 and 18, that is, the view direction corresponding blocks, are divided into partition blocks having various shapes, the motion vector prediction value MV _P can be obtained as follows.

(i) When the blocks 'TF0', 'TB0', 'VF0', or 'VB0' of FIG. 7 are divided into partition blocks having various shapes, the motion vector of the partition block located on the leftmost top of the block is determined. I use it.

(ii) When the blocks 'TF1', 'TB1', 'VF1', or 'VB1' of FIG. 7 are divided into partition blocks having various shapes, the motion vector of the partition block located at the top right of the block is determined. I use it.

(iii) When the blocks 'TF2', 'TB2', 'VF2', or 'VB2' of FIG. 7 are divided into partition blocks having various shapes, the motion vector of the partition block located at the bottom left of the block is determined. I use it.

(iv) When the blocks 'TF3', 'TB3', 'VF3', or 'VB3' of FIG. 7 are divided into partition blocks having various shapes, the motion vector of the partition block located on the leftmost top of the block is determined. I use it.

(v) When a block 'TF4', 'TB4', 'VF4', or 'VB4' of FIG. 7 is divided into partitions having various shapes, the motion vector of the partition block located at the upper left of the block is used. .

(vi) When the blocks 'TF5', 'TB5', 'VF5', or 'VB5' of FIG. 7 are divided into partition blocks having various shapes, the motion vector of the partition block located at the bottom rightmost side of the block Use

(vii) When the blocks 'TF6', 'TB6', 'VF6', or 'VB6' of FIG. 7 are divided into partitions of various shapes, the motion vector of the partition block located at the bottom left of the block is used. .

(viii) When the blocks 'TF7', 'TB7', 'VF7', or 'VB7' of FIG. 7 are divided into partition blocks having various shapes, the motion vector of the partition block located at the top right of the block is determined. I use it.

(ix) When the block 'TF8', 'TB8', 'VF8', or 'VB8' of FIG. 7 is divided into blocks having various shapes, the motion vector of the partition block located on the leftmost top of the block Use

For example, the above-described subset S ₁ = {'A', 'B', 'C', 'D', 'VF0', 'is a set of blocks that can be used to obtain a motion vector prediction value MV _P. VB0 ',' TF0 ', and' TB0 '} are selected, and each block constituting the subset S ₁ is divided into partition blocks of various shapes. In this case, the above-mentioned A- (i) is applied to the block 'A', the above-mentioned A- (ii) is applied to the block 'B', and the above-mentioned A to the block 'C'. -(iii) is applied, and in the case of block 'D', the aforementioned A- (iv) is applied, and in the case of blocks 'VF0''VB0','TF0', and 'TB0', B- ( By applying i), the motion vector prediction value MV _P can be obtained using all or a part of the motion vectors of the partition block to be applied.

The current block is 8 by 16 pixels

As shown in FIG. 5A, when the current block has an 8 × 16 pixel size, a block used to obtain a motion vector prediction value MV _P may be divided into an 8 × 16 left block and an 8 × 16 right block. .

(A) The set of blocks that can be used to obtain the motion vector prediction value MV _P of the 8 × 16 left block may be limited to the next set S ₅ , which is another subset of the set S.

(i) The set S ₅ may be limited to include only block 'A' (see FIG. 7), which is a neighboring block. In this case, the motion vector of the block 'A' is directly used as the motion vector prediction value MV _P.

(ii) As another example, when the current block has a motion vector in the view direction, the set S ₅ is a block 'A', which is a neighboring block, and blocks 'TF0', 'TF1', and 'TF2', which are time-direction corresponding blocks. , 'TF4', 'TF5', 'TF7', 'TB0', 'TB1', 'TB2', 'TB4', 'TB5', and 'TB7' (see FIG. 7). In this case, the set S ₅ = {'A', 'TF0', 'TF1', 'TF2', 'TF4', 'TF5', 'TF7', 'TB0', 'TB1', 'TB2', ' TB4 ',' TB5 ',' and the TB7 '}, using all or a portion of the motion vectors of the blocks belonging to the set S ₅ can obtain the motion vector prediction value (MV _P). For example, the median value of the motion vectors of the blocks 'TF5', 'TF1', 'TF7', 'TB5', 'TB1', and 'TB7', which are one of the subsets of the set S ₅ , is obtained. MV _P ). In addition, the median value of the motion vectors of the blocks 'A', 'TF1', and 'TB1' which are other subsets of the set S ₅ may be obtained, and this may be used as the motion vector prediction value MV _P.

(iii) As another example, when the current block has a motion vector in the time direction, the set S ₅ is a block 'A' which is a neighboring block and blocks 'VF0', 'VF1', and 'VF2' which are view direction corresponding blocks. , 'VF4', 'VF5', 'VF7', 'VB0', 'VB1', 'VB2', 'VB4', 'VB5', and 'VB7' (see FIG. 7). In this case, the set S ₅ = {'A', 'VF0', 'VF1', 'VF2', 'VF4', 'VF5', 'VF7', 'VB0', 'VB1', 'VB2', ' VB4 ',' VB5 ',' and the VB7 '}, using all or a portion of the motion vectors of the blocks belonging to the set S ₅ can obtain the motion vector prediction value (MV _P). For example, the median value of the motion vectors MV of the blocks 'VF5', 'VF1', 'VF7', 'VB5', 'VB1', and 'VB7', which are one of the subsets of the set S ₅ , is obtained. It can be set as the vector prediction value MV _P. In addition, the median value of the motion vectors of the blocks 'A', 'VF1', and 'VB1' which are other subsets of the set S ₅ may be obtained, and this may be used as the motion vector prediction value MV _P.

(B) The set of blocks that can be used to obtain the motion vector prediction value MV _P of the 8 × 16 right block may be limited to the next set S ₆ , which is another subset of the set S.

(i) The set S ₆ may be limited to include only the block 'C' (see FIG. 7) which is a reference block. In this case, the motion vector of the block 'C' is directly used as the motion vector prediction value MV _P.

(ii) As another example, when the current block has a motion vector in the view direction, the set S ₆ is a block 'C' which is a neighboring block and blocks 'TF0', 'TF2', 'TF3', It may be limited to include 'TF4', 'TF6', 'TF8', 'TB0', 'TB2', 'TB3', 'TB4', 'TB6', and 'TB8'. In this case, the set S ₆ = {'C', 'TF0', 'TF2', 'TF3', 'TF4', 'TF6', 'TF8', 'TB0', 'TB2', 'TB3', ' TB4 ',' TB6 ',' is the TB8 '}, using all or a portion of the motion vectors of the blocks belonging to the set S ₆ can obtain the motion vector prediction value (MV _P). For example, the median of the motion vectors of the blocks 'TF6', 'TF3', 'TF8', 'TB6', 'TB3', and 'TB8', which are one of the subsets of the set S ₆ , is obtained, and the motion vector predicted value ( MV _P ). In addition, the median value of the motion vectors of the blocks 'C', 'TF3', and 'TB3', which are other subsets of the set S ₆ , may be obtained and may be used as the motion vector prediction value MV _P.

(iii) As another example, when the current block has a motion vector in the time direction, the set S ₆ is a block 'C' which is a neighboring block and blocks 'VF0', 'VF2', and 'VF3' which are view direction corresponding blocks. , 'VF4', 'VF6', 'VF8', 'VB0', 'VB2', 'VB3', 'VB4', 'VB6', and 'VB8' (see FIG. 7). In this case, the set S ₆ = {'C', 'VF0', 'VF2', 'VF3', 'VF4', 'VF6', 'VF8', 'VB0', 'VB2', 'VB3', ' VB4 ',' VB6 ',' and the VB8 '}, using all or a portion of the motion vectors of the blocks belonging to the set S ₆ can obtain the motion vector prediction value (MV _P). For example, the median value of the motion vectors of the blocks 'VF6', 'VF3', 'VF8', 'VB6', 'VB3', and 'VB8', which are one of the subsets of the set S ₆ , is obtained. MV _P ). In addition, the median value of the motion vectors of the blocks 'C', 'VF3', and 'VB3', which are other subsets of the set S ₆ , may be obtained and may be used as the motion vector prediction value MV _P.

The current block is 16 by 8 pixels

In the case where the current block has a 16 × 8 pixel size as shown in FIG. 5B, available blocks for obtaining the motion vector prediction value MV _P are the 16 × 8 top block and the 16 × 8 bottom block. You can think about the case separately.

(A) The set of blocks that can be used to obtain the motion vector prediction value MV _P of the 16 × 8 upper block may be limited to the next set S ₇ which is another subset of the set S.

(i) The set S ₇ may be limited to include only block 'B' (see FIG. 7), which is a neighboring block. In this case, the motion vector of the block 'B' is directly used as the motion vector prediction value MV _P.

(ii) As another example, when the motion vector MV _C of the current block refers to an image in the view direction, the set S ₇ is a block 'B', which is a neighboring block, and blocks' TF0 ',', which are corresponding blocks in a time direction. It may be limited to include TF1, 'TF2', 'TF3', 'TF5', 'TF6', 'TB0', 'TB1', 'TB2', 'TB3', 'TB5', and 'TB6'. . In this case, the set S ₇ = {'B', 'TF0', 'TF1', 'TF2', 'TF3', 'TF5', 'TF6', 'TB0', 'TB1', 'TB2', ' TB3 ',' TB5 ',' and the TB6 '}, using all or a portion of the motion vectors of the blocks belonging to the set S ₇ can obtain the motion vector prediction value (MV _P). For example, the median of the motion vectors of the blocks 'TF5', 'TF2', 'TF6', 'TB5', 'TB2', and 'TB6', which are one of the subsets of the set S ₇ , is obtained, and the motion vector predicted value ( MV _P ). In addition, the median value of the motion vectors of the blocks 'B', 'TF2', and 'TB2' which are other subsets of the set S ₇ may be obtained, and this may be used as the motion vector prediction value MV _P.

(iii) As another example, when the motion vector MV _C of the current block refers to an image in the time direction, the set S ₇ is a block 'VF0', 'VF1' which is a neighboring block 'B' and a view direction corresponding block. And 'VF2', 'VF3', 'VF5', 'VF6', 'VB0', 'VB1', 'VB2', 'VB3', 'VB5', and 'VB6'. In this case, the set S ₇ = {'B', 'VF0', 'VF1', 'VF2', 'VF3', 'VF5', 'VF6', 'VB0', 'VB1', 'VB2', ' VB3 ',' VB5 ',' and the VB6 '}, using all or a portion of the motion vectors of the blocks belonging to the set S ₇ can obtain the motion vector prediction value (MV _P). For example, the median value of the motion vectors of the blocks 'VF5', 'VF2', 'VF6', 'VB5', 'VB2', and 'VB6', which are one of the subsets of the set S ₇ , is obtained. MV _P ). In addition, the median value of the motion vectors of blocks 'B', 'VF2', and 'VB2', which are other subsets of the set S ₇ , may be obtained, and the motion vector predicted value MV _P may be obtained.

(B) The set of blocks that can be used to obtain the motion vector prediction value MV _P of the 16 × 8 lower block may be limited to the next set S ₈ which is another subset of the set S.

(i) The set S ₈ may be limited to include only block 'A' which is a neighboring block. In this case, the motion vector of the block 'A' is directly used as the motion vector prediction value MV _P.

(ii) As another example, when the motion vector MV _C of the current block refers to an image in the view direction, the set S ₈ is a block 'A', which is a neighboring block, and blocks' TF0 ',', which are corresponding blocks in a time direction. It may be limited to include TF1 ',' TF3 ',' TF4 ',' TF7 ',' TF8 ',' TB0 ',' TB1 ',' TB3 ',' TB4 ',' TB7 ', and' TB8 '. . In this case, the set S ₆ = {'A', 'TF0', 'TF1', 'TF3', 'TF4', 'TF7', 'TF8', 'TB0', 'TB1', 'TB3', ' TB4 ',' TB7 ', and' TB8 '}, and the motion vector prediction value MV _P can be obtained using the motion vectors of all or some of the blocks belonging to the set S ₈ . For example, the median value of the motion vectors of the blocks 'TF7', 'TF4', 'TF8', 'TB7', 'TB4', and 'TB8', which are one of the subsets of the set S ₈ , is obtained. MV _P ). In addition, the median value of the motion vectors of the blocks 'A', 'TF4', and 'TB4' which are other subsets of the set S ₈ may be obtained, and this may be used as the motion vector prediction value MV _P.

(iii) As another example, when the motion vector MV _C of the current block refers to an image in the time direction, the set S ₈ is a block 'A' that is a neighboring block, a block 'VF0' that is a view-direction corresponding block, Can be limited to include 'VF1', 'VF3', 'VF4', 'VF7', 'VF8', 'VB0', 'VB1', 'VB3', 'VB4', 'VB7', and 'VB8' have. In this case, the set S ₈ = {'A', 'VF0', 'VF1', 'VF3', 'VF4', 'VF7', 'VF8', 'VB0', 'VB1', 'VB3', ' VB4 ',' VB7 ', and' VB8 '}, and the motion vector prediction value MV _P can be obtained using the motion vectors of all or some of the blocks belonging to the set S ₈ . For example, the median value of the motion vectors of the blocks 'VF7', 'VF4', 'VF8', 'VB7', 'VB4', and 'VB8' which are one of the subsets of the set S ₈ is obtained, and the motion vector predicted value ( MV _P ). In addition, the median value of the motion vectors of the blocks 'A', 'VF4', and 'VB4', which are other subsets of the set S ₈ , may be obtained and may be used as the motion vector prediction value MV _P.

When considering the distance of the reference image

In the inter prediction encoding, a motion vector prediction value (MV _P ) is adaptively determined according to a distance from a reference picture to which a current picture (current block) refers in a time direction or a view direction, that is, a time direction reference picture or a view direction reference picture. You can select a block to use. This is to obtain the motion vector prediction value MV _P of the current block. When the distance to the reference picture referred to by the current picture is long, the motion vector MV _C of the current block and the motion vector MV of the corresponding block in time direction are obtained. _T ) or the motion vector (MV _V ) of the view direction corresponding block are highly likely to have a large difference or different characteristics, so that the motion vector prediction value MV _P may not be accurate. Therefore, as the distance from the current picture to the reference picture increases, the number of reference blocks used to obtain the motion vector prediction value MV _P is reduced or the prediction from the reference picture is not allowed, thereby increasing the prediction coding efficiency. Can be improved. This will be described in more detail below.

Assume that the motion vector MV _C of the current block refers to an image in the view direction. In this case, the set of time direction corresponding blocks used to obtain the motion vector prediction value MV _P according to the distance from the current image to the temporal reference image can be more specifically defined.

For example, when the distance from the current image to the temporal reference image is 1, reference blocks that can be used to obtain a motion vector prediction value MV _P are illustrated with neighboring blocks corresponding to reference numeral 10 of FIG. 7. Any time direction corresponding block corresponding to the reference numerals 12 and 14 of 7 may be limited. That is, a set of blocks that can be used to obtain a motion vector prediction value MV _P is a set S ₉ = {'A', 'B', 'C', 'D', 'TF0' which is a subset of the set S. , 'TF1', 'TF2', 'TF3', 'TF4', 'TF5', 'TF6', 'TF7', 'TF8', 'TB0', 'TB1', 'TB2', 'TB3', ' TB4 ',' TB5 ',' TB6 ',' TB7 ',' TB8 '}. The motion vector prediction value MV _P is obtained using the motion vectors of all or some of the blocks belonging to the set S ₉ .

When the distance from the current image to the temporal reference image is 2, blocks that can be used to obtain a motion vector prediction value MV _{P are} shown in the neighboring blocks of reference numeral 10 and reference numerals 12 and 14 of FIG. It can be further limited to blocks 'TF0' and 'TB0', which are blocks in the same position as the current block. That is, a set of blocks that can be used to obtain a motion vector prediction value MV _P is set as another subset of the set S, set S ₁₀ = {'A', 'B', 'C', 'D', 'TF0. ',' TB0 '}. The motion vector prediction value MV _P is obtained using the motion vectors of all or part of the blocks belonging to the set S ₁₀ .

In addition, when the distance from the current image to the temporal reference image is 3 or more, the motion vector predicted value (MV _N ) of the neighboring block is not used without using the motion vector MV _T of the corresponding block in the time direction. MV _P ) can be obtained.

Table 1 summarizes the aforementioned examples of obtaining a motion vector prediction value MV _P by adaptively selecting a block according to the distance from the current image to the temporal reference image.

Distance to temporal reference image (D) Set of reference blocks that can be used to obtain the motion vector prediction value (MV _P ) D = 1 'A', 'B', 'C', 'D', 'TF0', 'TF1', 'TF2', 'TF3', 'TF4', 'TF5', 'TF6', 'TF7', 'TF8 '' TB0 ',' TB1 ',' TB2 ',' TB3 ',' TB4 ',' TB5 ',' TB6 ',' TB7 ',' TB8 ' D = 2 'A', 'B', 'C', 'D', 'TF0', 'TB0' 3 < D 'A', 'B', 'C', 'D'

As another example, assume that the motion vector MV _C of the current block refers to an image in a time direction. In this case, the set of blocks used to obtain the motion vector prediction value MV _P can be more specifically defined according to the distance from the current image to the view direction reference image.

For example, when the distance from the current image to the view direction reference image is 1, a block that can be used to obtain a motion vector prediction value MV _P and neighboring blocks corresponding to reference numeral 10 of FIG. It may be limited to the view direction corresponding blocks corresponding to the reference numerals 16 and 18. That is, a set of blocks that can be used to obtain a motion vector prediction value MV _P is set to another subset of the set S, set S ₁₁ = {'A', 'B', 'C', 'D', 'VF0'. ',' VF1 ',' VF2 ',' VF3 ',' VF4 ',' VF5 ',' VF6 ',' VF7 ',' VF8 ',' VB0 ',' VB1 ',' VB2 ',' VB3 ', 'VB4', 'VB5', 'VB6', 'VB7', 'VB8'}. The motion vector prediction value MV _P is obtained using the motion vectors of all or part of the blocks belonging to the set S ₁₁ .

When the distance from the current image to the view direction reference image is 2, a block that can be used to obtain a motion vector prediction value MV _P is obtained from neighboring blocks at reference numeral 10 of FIG. 7 and reference numerals 16 and 18. It may be further limited to blocks 'VF0' and 'VB0' which are view direction corresponding blocks at the same position as the current block. That is, a set of blocks that can be used to obtain a motion vector prediction value MV _P is set to another subset of the set S, set S ₁₂ = {'A', 'B', 'C', 'D', 'VF0'. ',' VB0 '}. The motion vector prediction value MV _P is obtained using the motion vectors of all or some of the blocks belonging to the set S ₁₂ .

In addition, when the distance from the current image to the view direction reference image is 3 or more, the motion vector prediction value using only the motion vector MV _N of neighboring blocks without using the motion vector MV _V of the corresponding view direction block. (MV _P ) can be obtained.

Table 2 summarizes the aforementioned examples of obtaining a motion vector prediction value MV _P by adaptively selecting a block according to the distance from the current image to the view direction reference image.

Distance to view orientation reference image (D) Set of reference blocks that can be used to obtain the motion vector prediction value (MV _P ) D = 1 'A', 'B', 'C', 'D', 'VF0', 'VF1', 'VF2', 'VF3', 'VF4', 'VF5', 'VF6', 'VF7', 'VF8 '' VB0 ',' VB1 ',' VB2 ',' VB3 ',' VB4 ',' VB5 ',' VB6 ',' VB7 ',' VB8 ' D = 2 'A', 'B', 'C', 'D', 'VF0', 'VB0' 3 < D 'A', 'B', 'C', 'D'

Syntax for Predicting Motion Vectors

As described above, in multi-view video coding according to the present invention, when the motion vector prediction value MV _P of the current block is obtained, (1) the motion vector MV _N of the neighboring block, and (2) the 'time direction reference image' Uses the motion vector MV _T of the 'time direction corresponding block', and / or (3) the motion vector MV _V of the 'view direction reference block' in the 'view direction reference image'. The motion vector prediction method in multi-view video coding according to the present invention uses a sequence header, a slice header, and a variable length code, a fixed length code, and / or an Exp-Golomb code. / Or appropriately represented by a macroblock header.

According to a preferred embodiment of the present invention, in addition to obtaining a motion vector prediction value MV _P using only the existing method defined in H.264 / AVC, that is, the motion vector MV _N of the above (1), The motion vector MV _T of (2) and / or the motion vector MV _V of (3) are also used together. In such a case, using the motion vector (MV _N ) of (1) can be expressed in the same way as defined in H.264 / AVC. Therefore, in this case, it is only necessary to additionally express whether to use the motion vector MV _T of (2) and / or the motion vector MV _V of (3). Hereinafter, a description will be given of a method of expressing in various syntaxes whether to use the motion vector MV _T of (2) and / or the motion vector MV _V of (3). However, the method of expressing the use of the motion vector MV _N of (1) in the following example is not limited to the existing method defined in H.264 / AVC.

(i) Slice information (code) about whether to use the motion vector MV _T of the time-direction corresponding block and / or the motion vector MV _V of the view-direction corresponding block when obtaining the motion vector prediction value MV _P. Can be represented in a header. For example, when information on whether to use the motion vector MV _T of the time direction corresponding block is represented by flag information in the slice header, when the flag becomes '1', the time direction corresponding block When the motion vector MV _T is used and the flag is '0', the motion vector MV _T of the reference block in the time direction may not be used.

In the above-described example, the motion vector MV _C of the current block refers to an image in the view direction, and the motion vector MV _{N of} neighboring blocks is not available, or among the motion vectors MV _N of the neighboring blocks. When there is no reference to an image, the coding efficiency can be improved by obtaining a motion vector prediction value MV _P using the motion vector MV _T of the corresponding block in the time direction. In this case, flag information indicating that the time direction corresponding block is used is included in a slice header.

However, if there is a scene change in the time direction or a sudden change in brightness occurs, the motion vector MV _T of the corresponding time direction block is likely to have a completely different value from the motion vector MV _C of the current block. . In this case, even when the motion vector MV _T of the corresponding time direction block refers to an image in the same view direction as the motion vector MV _C of the current block, the time in obtaining the motion vector prediction value MV _P is obtained. It is preferable not to use the motion vector MV _T of the direction correspondence block. That is, even when there is a scene change in the time direction or a sudden change in brightness occurs, the motion vector prediction value MV _P is obtained using the motion vector MV _T of the corresponding time direction block, and the motion vector is obtained. When the motion vector difference value MV _D is obtained from the prediction value MV _P , the information amount of the motion vector difference value MV _D may increase, and thus, the efficiency of predictive coding may be deteriorated. Therefore, in such a case, it is preferable to include flag information that the reference block in the time direction is not used in the slice header.

As described above, when the motion vector prediction value MV _P is obtained, information about whether to use the motion vector MV _T of the corresponding time direction block is expressed in a slice header to further improve prediction coding efficiency. There is a number. This principle is equally applicable to the motion vector MV _V of the view direction corresponding block.

Table 3 illustrates an example of syntax including flag information indicating whether to use a motion vector MV _T of a time direction corresponding block in a slice header. Referring to Table 3, the syntax for the slice head corresponds to a flag "use_list0_motion_prediction_flag" and a list 1 (list1), which are flags indicating whether to perform prediction on a temporal reference image corresponding to list0. And "use_list1_motion_prediction_flag" which is a flag indicating whether prediction is performed on the temporal direction reference image. Here, "use_list0_motion_prediction_flag" corresponds to a P-picture or a B-picture, and "use_list1_motion_prediction_flag" corresponds only to a B-picture. Table 3 is equally applicable to a syntax including flag information indicating whether to use a motion vector (MV _V ) of a view direction corresponding block in a slice header.

slice_header () { C DescripTor first_mb_in_slice 2 ue (T) slice_type 2 ue (T) pic_parameter_set_id 2 ue (T) frame_num 2 u (T) … if ((profile = = MULTI_VIEW_PROFILE)
&& slice_Type! = I && slice_Type! = SI) { use_list0_motion_prediction_flag 2 u (1) if (slice_type = = B) { use_list1_motion_prediction_flag 2 u (1) } } … }

As another example, the meaning of each flag defined in Table 3 may be used differently. For example, "use_list0_motion_prediction_flag" is used as a flag indicating whether to perform motion vector prediction using a 'time direction reference image' temporally before the current image in obtaining a motion vector prediction value MV _P. use_list1_motion_prediction_flag "may be used as a flag representing whether or not according to obtain the motion vector prediction value (MV _P), using a" time reference picture direction, subsequent in time than the current picture performs motion vector prediction.

As another embodiment, when the motion vector prediction value MV _P expressed in the slice header is obtained, the motion vector MV _T of the corresponding time direction block and / or the motion vector MV _V of the view direction corresponding block are used. Information on whether or not may be expressed using the Exp-Golomb code, variable length code, or fixed length code used in H.264 / AVC instead of the flag information as defined in Table 3.

In the above-described embodiments, whether the motion vector MV _T of the time direction corresponding blocks or the motion vector MV _V of the view direction corresponding blocks in slice units are used to obtain the motion vector prediction value MV _P is various. It can be decided by the method. As an example of such a determination method, an average value of luminance values of pixels may be used in units of slices. For example, when determining whether to use the motion vectors MV _T of time-corresponding blocks, an average value of luminance values of each pixel is obtained in slice units in the current image and the time direction reference image, and the difference in the average value is predetermined. If the value of α is greater than or equal to, the motion vector MV _T of the corresponding blocks in time direction is not used. On the contrary, if the difference in the average value is less than or equal to the predetermined value α, the motion of the corresponding blocks in time direction is used. One may decide to use the vector MV _T.

(ii) when obtaining the motion vector prediction value MV _P , sequence information on whether to use the motion vector MV _T of the corresponding time direction block and / or the motion vector MV _V of the corresponding view direction block. It can also be expressed in the syntax for the header. In some embodiments, information on whether to use the motion vector MV _T of the corresponding time direction block and / or the motion vector MV _V of the corresponding view direction block may be expressed in the syntax for the macroblock. have.

(iii) The methods of (i) and (ii) described above may be used together, or only one of them may be used.

Motion vectors (MV) of the selected block _N , MV _T , And / or MV _V Motion vector prediction (MV) _P How to get

A motion vector prediction value (MV _P) not less than the two used for the selected motion vector to obtain one, using the selected motion vector to obtain the motion vector prediction value (MV _P) has a number of reasons. For example, the median of the selected motion vectors can be obtained as specified in H.264 / AVC, and this intermediate value can be used as the motion vector prediction value MV _P. And it may be an average value of the selected motion vector as a motion vector prediction value (MV _P) or the motion vectors are also selected by the motion vector prediction value to the smallest difference from the block of motion vectors from (MV _P). Hereinafter, a case of selecting N motion vectors and obtaining an intermediate value of the selected motion vectors will be described in detail.

(i) As an example of selecting N blocks for selecting N motion vectors from the aforementioned set S or a subset thereof, such as one of the sets S ₁ to S ₁₂ , the blocks belonging to the set S You can set the priority between them (which can be absolute or relative) and select N blocks with that priority. The priority may have any order. For example, among the neighboring blocks (blocks corresponding to reference numeral 10 of FIG. 7), priority may be determined in the order of blocks 'A', 'B', 'C', and 'D'.

In addition, the priority may be determined according to the position of the block corresponding to the time direction block or the view direction block. For example, assign a high priority to a block at the same position as the current block, assign a next higher priority to a block located in the horizontal or vertical direction with respect to a block at the same position as the current block, The lowest priority may be assigned to blocks located diagonally with respect to blocks of the same position. Referring to FIG. 7, the highest priority is assigned to the block 'TF0', which is a block in the same position as the current block, and the block 'TF1', which is a block located in the horizontal and vertical directions with respect to the block 'TF0'. , 'TF2', 'TF3', and 'TF4' are assigned the next highest priority, and the blocks 'TF5', 'TF6', 'TF7', And 'TF8' may be assigned the lowest priority. And blocks 'TF1', 'TF2', 'TF3', and 'TF4' and blocks 'TF5', 'TF6', 'TF7', and 'TF8' respectively assign priorities in random order or in the order listed. Priority can be assigned.

(ii) When the number of blocks available for obtaining the motion vector prediction value MV _P in the example of (i) is smaller than N, for example, when M (1 <M <N), The median value can be obtained using the motion vector, and the median value can be used as the motion vector prediction value MV _P. Alternatively, in the above-described example, the NM motion vectors MV may be set to (0.0, 0.0), and the motion vector prediction value MV _P may be obtained by performing an intermediate value operation on the N motion vectors.

(iii) In the example of (i), when the number of blocks that can be used to obtain the motion vector prediction value MV _P is one, for example, the motion vector MV of the one block is determined by the motion vector prediction value MV _P. You can also do

(iv) In the example of (i), if there is no block available for obtaining the motion vector prediction value MV _P , N in accordance with the priority from the motion vector MV of the neighboring block as in the conventional method. Blocks can be selected, and the median value of the motion vectors MV of the selected N blocks can be set as the motion vector prediction value MV _P. As another example, the motion vector prediction value MV _P may be set as (0.0, 0.0) without obtaining an intermediate value.

FIG. 13 is a flowchart illustrating a method of obtaining a motion vector prediction value MV _P by performing an intermediate value operation on N or less motion vectors according to the above-described embodiment. Hereinafter, the flowchart will be described in detail with reference to FIG. 13.

First, as the set S or a subset thereof, calculate the number M of blocks that can be used to obtain a motion vector prediction value MV _P according to one or a combination of the above-described embodiments of the present invention. (S31).

The number M of blocks is compared with the size N (S32). As a result of the determination, if M <N, branching is made to step S34, and if M? N, branching to step S33.

In step S33, N blocks are selected according to the priority among the M blocks, and an intermediate value for motion vectors of the selected N blocks is obtained (S33). The median value thus obtained becomes the motion vector prediction value MV _P for the current block, in which case the procedure ends here.

In step S34, it is determined whether M is 0 (S34). As a result of the determination, if M is not 0, the flow branches to step S36. If M is 0, the flow branches to step S35.

In step S35, N blocks are selected from neighboring blocks, and an intermediate value for motion vectors of the selected block is obtained (S35). The median value thus obtained becomes the motion vector prediction value MV _P for the current block, in which case the procedure ends here.

In step S36, it is determined whether M is 1 (S36). As a result of the determination, if M is 1, the flow branches to step S38. If M is not 1, the flow branches to step S37.

In step S37, the (NM) motion vectors (MV) are set to (0.0, 0.0), and the median values for the motion vectors (MV) of the M blocks and (NM) motion vectors set to (0.0, 0.0). Obtain (S37). The median value thus obtained becomes the motion vector prediction value MV _P for the current block, in which case the procedure ends here.

Since M is 1 in step S38, the motion vector MV of the block is set as the motion vector predicted value MV _P , and the procedure ends (S38).

According to an embodiment of the present invention, the motion vector prediction value MV _P may be obtained without using a time direction block or a view direction block. In this case, as in the conventional method specified in H.264 / AVC, the motion vector prediction value MV _P is obtained using only neighboring blocks of the current block. That is, the motion vector prediction value MV _P is obtained by performing an intermediate value operation on the motion vectors MV _N of blocks 'A', 'B', and 'C' corresponding to reference numeral 10 of FIG. Alternatively, when the motion vector MV of the block 'C' does not exist, the block 'D' is used instead of the block 'C', and the motion vectors MV of the blocks 'A', 'B', and 'D' are used. _The motion vector prediction value (MV _P ) is obtained through the median operation on _N ).

Motion vector prediction (MV _P Application example of how to obtain

Combining the above-described methods, a variety of methods for obtaining the motion vector prediction value MV _P can be made. Hereinafter, one method thereof will be described with reference to FIG. 14.

14 is a flowchart illustrating a method of obtaining a motion vector prediction value MV _P according to an embodiment of the present invention.

Referring to FIG. 14, first, a distance D in a time direction from a current image to an image (a time direction reference image) referenced by the current image in the time direction is calculated (S501). Here, the distance D in the time direction may mean, for example, a difference between frames of the current image and the reference image. For example, if the current image is an image of T2 time and the reference image is an image of T0 time, the distance D in the time direction becomes two.

Then, it is determined whether or not the distance D in the time direction is 2 or less (S502). As a result of the determination, if D is 2 or less, the flow branches to step S503. If D is larger than 2, the flow branches to step S517.

In operation S503, it is determined whether there is a reference to the same image as the image referenced by the motion vector MV _C of the current block among the motion vectors MV _N of the neighboring blocks. As a result of the determination, if there is a block referring to the same image as the image referred to by the motion vector MV _C of the current block among the motion vectors MV _N of the neighboring blocks, the flow branches to step S517; do.

In operation S504, it is determined whether the motion vector MV _C of the current block refers to a view direction reference image or a time direction reference image (S504). As a result of the determination, if the motion vector MV _C of the current block refers to the view direction reference image, it branches to step S505, and if it refers to the time direction reference image, it branches to step S511.

In operation S505, among the motion vectors MV _T of the corresponding blocks in the time direction of the temporal reference image, the number of blocks N _T that refers to the image of the same view as the view referenced by the motion vector MV _C of the current block. ) Determines whether 0 is 0 (S505). As a result of the determination, if the number N _T of blocks referring to the image of the same view is not 0, the process branches to step S506, and if it is 0, the process branches to step S517.

In operation S506, it is determined whether the number N _T of corresponding temporal direction blocks referring to the image of the same view as the view referenced by the motion vector MV _C of the current block is less than 3 (S506). As a result of the determination, if the number N _T of time-direction corresponding blocks referring to the image of the same view is less than 3, the process branches to step S508, and if more than 3, the process branches to step S507.

In operation S507, three motion vectors are selected according to the priority among the motion vectors MV _T of the corresponding blocks in the time direction referring to the same view as the view referred to by the motion vector MV _C of the current block, and the selected 3 is selected. A motion vector prediction value MV _P is obtained by performing an intermediate value operation on the two motion vectors MV _T (S507). When the motion vector prediction value MV _P is obtained in this manner, the procedure ends.

In operation S508, it is determined whether the number N _T of time-corresponding blocks referring to the same view as the view referenced by the motion vector MV _C of the current block is 1 (S508). As a result of the determination, when the number N _T of time-direction corresponding blocks referring to the same view is 1, the process branches to step S510. When the number N _T of time-direction corresponding blocks referring to the same view is 2, the step is performed. Branch to S509.

In operation S509, intermediate values of the motion vectors MV _T and the motion vectors (0.0, 0.0) of the corresponding block in the direction of time referencing the same view as the view referenced by the motion vector MV _C of the current block may be obtained. The motion vector prediction value MV _P is obtained (S509). When the motion vector prediction value MV _P is obtained in this manner, the procedure ends.

In operation S510, the motion vector MV _T of the corresponding block in the direction of time referring to the same view as the view referenced by the motion vector MV _C of the current block is a motion vector prediction value MV _P (S510). When the motion vector prediction value MV _P is obtained in this manner, the procedure ends.

In operation S511, among the motion vectors MV _V of the view direction corresponding blocks of the view direction reference image, the number of blocks (N _V ) referring to the image having the same time zone as the time zone referred to by the motion vector MV _C of the current block is referred to. ) Is determined to be 0 (S511). As a result of the determination, if the number N _V of blocks referring to the image of the same time zone is not 0, the process branches to step S512, and if it is 0, the process branches to step S517.

In operation S512, it is determined whether the number N _V of view-direction corresponding blocks referring to the image of the same time zone as the time zone referenced by the motion vector MV _C of the current block is less than 3 (S512). As a result of the determination, if the number N _V of view-direction corresponding blocks referring to the image of the same time zone is less than 3, the process branches to step S514, and if more than 3, the process branches to step S513.

In operation S513, three motion vectors are selected from among the motion vectors MV _V of the view direction correspondence block referring to the same time zone that the motion vector MV _C of the current block refers to, and the selected 3 is selected. The motion vector prediction value MV _P is obtained by performing an intermediate value operation on the two motion vectors MV _V (S513). When the motion vector prediction value MV _P is obtained in this manner, the procedure ends.

In operation S514, it is determined whether the number N _V of view-direction corresponding blocks referring to the same time zone as the time zone referred to by the motion vector MV _C of the current block is 1 (S514). As a result of the determination, when the number N _V of the view direction correspondence blocks referring to the same time zone is 1, the process branches to step S516, and when the number N _V of the view direction correspondence blocks referring to the same time zone is 2, the step is performed. Branches to S515.

In operation S515, the median operation is performed on the motion vectors MV _V and the motion vectors (0.0, 0.0) of the view direction corresponding block referring to the same time zone as the time zone referenced by the motion vector MV _C of the current block. The motion vector prediction value MV _P is obtained (S515). When the motion vector prediction value MV _P is obtained in this manner, the procedure ends.

In operation S516, the motion vector MV _V of the view direction corresponding block that refers to the same time zone that the motion vector MV _C of the current block refers to is the motion vector prediction value MV _P (S516). When the motion vector prediction value MV _P is obtained in this manner, the procedure ends.

Finally, in step S517, a motion vector prediction value MV _P is obtained using only the same method as the existing method specified in H.264 / AVC, that is, only motion vectors MV _N of neighboring blocks of the current block (S517). . When the motion vector prediction value MV _P is obtained in this manner, the procedure ends.

As a modified embodiment or more specific embodiment of the present invention described with reference to FIG. 14, one or more of the following methods may be additionally used.

1) When obtaining the motion vector prediction value MV _P , flag information indicating whether to use the motion vector MV _T of the time direction corresponding block or the motion vector MV _V of the view direction corresponding block is added to the syntax of the sequence head. Can be included. For example, when using the motion vector MV _T of the time direction corresponding block and / or the motion vector MV _V of the view direction corresponding block, the flag is assigned to the sign '1', and if not used, the sign '0' is used. 'And vice versa.

2) When obtaining the motion vector prediction value MV _P , flag information indicating whether to use the motion vector MV _T of the time direction corresponding block or the motion vector MV _V of the view direction corresponding block is added to the syntax of the slice head. Can be included. For example, when using the motion vector MV _T of the time direction corresponding block and / or the motion vector MV _V of the view direction corresponding block, the flag is assigned to the sign '1', and if not used, the sign '0' is used. 'And vice versa.

3) When a block, for example, a neighboring block, a time direction corresponding block, or a view direction corresponding block, which is referred to to obtain a motion vector prediction value MV _P is divided into partition blocks having various sizes, partitions according to the positions of the blocks. The block can be selected appropriately. Referring to FIG. 7, the example of such a case will be described in more detail. The block 'A' uses the motion vector of the partition block located at the upper right, and the block 'B' uses the motion vector of the partition block located at the lower left. Block 'C' uses the motion vector of the partition block located at the lower left, block 'D' uses the motion vector of the partition block located at the lower right, and blocks 'TF0', 'TB0' and 'VF0'. ', And' VB0 'use the motion vector of the partition block located at the top left, and the blocks'TF1','TB1','VF1', and 'VB1' use the motion vector of the partition block located at the top right. , Blocks 'TF2', 'TB2', 'VF2', and 'VB2' use the motion vector of the partition block located at the lower left, and blocks 'TF3', 'TB3', 'VF3', and 'VB3' Using the motion vector of the upper left partition, the blocks 'TF4', 'TB4', 'VF4' and 'VB4' are on the upper left The motion vectors of one partition block are used, and the blocks 'TF5', 'TB5', 'VF5', and 'VB5' use the motion vectors of the partition block located at the lower right, and the blocks 'TF6', 'TB6', 'VF6' and 'VB6' use the motion vector of the partition block located at the lower left, and blocks 'TF7', 'TB7', 'VF7', and 'VB7' use the motion vector of the partition block located at the upper right. In addition, blocks 'TF8', 'TB8', 'VF8', and 'VB8' may use the motion vector of the partition block located in the upper left corner.

4) If the current block is a block having a size of 16 × 8 pixels, the above-described embodiment may be applied with a slight modification. For example, in the above-described steps S503 and S516, when the current block is the upper block of 16x8 pixel size, the neighboring block corresponds to block 'B' only, and if the current block is the lower block of 16x8 pixel size, The peripheral block may be modified to correspond to the block 'A' only, and the above-described embodiment may be applied as it is.

5) Even when the current block is a block having a size of 8x16 pixels, the above-described embodiment may be slightly modified and applied as it is. For example, in the above-described steps S503 and S516, when the current block is a left block of 8 × 16 pixels, the neighboring block corresponds only to block “A”, and when the current block is a right block of 8 × 16 pixels. The peripheral block may be modified to correspond to the block 'C' only, and the above-described embodiment may be applied as it is.

6) The priority between the time direction corresponding blocks and view direction corresponding blocks that can be used to obtain the motion vector prediction value MV _P is the same block as the current block, the block to the left of the block, The blocks may be in the order of blocks, blocks located on the right, blocks located on the bottom, blocks located on the upper left, blocks located on the upper right, blocks located on the lower left, and blocks located on the lower right.

The method of obtaining the motion vector prediction value MV _P according to the embodiment of the present invention is not limited to the above-described embodiment, and it is apparent to those skilled in the art that various various methods are possible.

Motion vector prediction (MV _P Coding apparatus using the method to obtain

15 is a block diagram illustrating a configuration of an apparatus for encoding a multiview image, according to an embodiment of the present invention. The encoding apparatus illustrated in FIG. 15 is a multiview image encoding apparatus using the method of obtaining the motion vector prediction value MV _P described above.

Referring to FIG. 15, the encoding apparatus includes a motion predictor 110, a motion compensator 120, a transform and quantizer 130, an entropy encoder 135, an inverse quantization and inverse changer 140, and a slice unit. Whether to use the motion vector MV _T of the time-direction corresponding block and / or the motion vector MV _V of the view-direction corresponding block (hereinafter, referred to as a 'use / determination unit'), and a motion vector predictor ( 160 and first, second, and third adders (102, 104, 106). Hereinafter, the configuration and operation of the encoding apparatus will be described in detail with reference to FIG. 15.

When data regarding the current image to be encoded as the inter image is input to the encoding apparatus, the motion predictor 110 performs motion prediction on a block basis to generate a motion vector (MV) of the current block. In the motion prediction, a reference picture of list0 and / or a reference picture of list1 is referred to. In the encoding of a multiview image, the reference image of each block may be a temporal reference image or a view direction reference image, so that the reference image of list0 includes the image and / or the current image at a time earlier than the current image. The image may be an image in front of the view, and the reference image of list 1 may be an image at a later time than the current image and / or an image in a rear view than the view to which the current image belongs. Therefore, the motion vector MV of the current block may refer to an image in a time direction or an image in a view direction.

The motion compensator 120 generates a motion compensated prediction image in units of blocks by using the motion vector MV generated by the motion predictor 110 and the reference image referenced by the motion predictor 110. The first adder 102 generates a differential image in units of blocks by dividing the predicted image input from the motion compensator 120 and the current image. The difference image output from the first adder 102 undergoes a transform and quantization process in the transform and quantizer 130, and the transform and quantizer 130 outputs a quantization coefficient. The quantization coefficients are entropy-encoded by the entropy encoder 135, and the entropy-encoded image information is formed into a bitstream along with other additional information and transmitted to a decoding apparatus or recorded in a storage medium. The quantization coefficients output from the transform and quantization unit 130 are reconstructed into differential images through inverse quantization and inverse transformation processes by the inverse quantization and inverse transform unit 140, and the reconstructed differential images are moved by the second adder 104. The reconstructed current image is generated by adding the predicted image output from the compensator 120. Since a series of processes from the motion compensator 120 to the second adder 104 are the same as those of the conventional video encoding apparatus, a detailed description thereof will be omitted herein.

In addition, the availability determining unit 150 uses the data of the current video, the reference video of list 0 (list0), and the reference video of list 1 (list1), to obtain a motion vector prediction value MV _P. It is determined whether the motion vector MV _T and / or the motion vector MV _V of the view-direction corresponding block are used, and information indicating the result is encoded and output. In this embodiment, this determination process is performed in units of slices. However, according to another embodiment of the present invention, this determination process may be performed in units of image sequence or blocks (eg, macroblocks). A block diagram showing an example of the configuration of the use determination unit 150 is illustrated in FIG. 16.

Referring to FIG. 16, the availability determining unit 150 includes an availability calculating unit 152 and an encoding unit 154 of availability information.

The available calculation unit 152 uses the data of the current video, the reference video of list 0 (list0), and the reference video of list 1 (list1), to obtain a motion vector prediction value MV _P. It is determined by calculating whether to use the motion vector MV _T and / or the motion vector MV _V of the view direction corresponding block. This determination may be performed in slice units, but is not limited thereto. And the specific process of performing this determination may use any of the methods described herein above.

The encoding unit 154 of the availability information is encoded by the available calculation unit 152 to encode a result of the motion vector MV _T of the corresponding block in the time direction and / or the motion vector MV _V of the corresponding block in the view direction. Slice information about whether or not to use The method of expressing the information in units of slices may use any of the methods described herein. For example, when the information of the slice unit is "1", MV _T or MV _V is used. If "0", MV _T and MV _V are not used, and only the motion vector MV _N of the neighboring block is used. Can be. The slice information may be included in a slice header and transmitted to the decoding apparatus or may be stored in a storage medium.

In addition, the motion vector predictor 160 includes a reference image of list 0, a reference image of list 1, and a reconstructed current image and a time direction corresponding block output from the availability selector 150. A motion vector prediction value MV _P is generated and output using slice information about whether to use the motion vector MV _T and / or the motion vector MV _V of the view direction corresponding block. 17 is a block diagram illustrating an example of the configuration of the motion vector predictor 160.

Referring to FIG. 17, the motion vector predictor 160 includes a block selector 162 and a motion vector predictor MV _P generator 164.

The block selector 162 is a slice unit relating to whether to use the motion vector MV _T of the time direction corresponding block and / or the motion vector MV _V of the view direction corresponding block output from the availability determination unit 150. A block that can be used to obtain a motion vector prediction value MV _P is selected based on the information of. This selection is based on blocks A, B, C, and D, which are the neighboring blocks of the current block in the reconstructed current image, and blocks XF0, XF1, XF2, and XF3, which are time-direction corresponding blocks and / or view-direction corresponding blocks in the reference image of list 0. , XF4, XF5, XF6, XF7, and XF8 (where X may be T or V) and block XB0, XB1, XB2, XB3, which are time-direction corresponding blocks and / or view-direction corresponding blocks in the reference image of List1 , XB4, XB5, XB6, XB7, and XB8. For example, when the information in units of slices is "1", a time direction corresponding block and / or view direction corresponding block are also selected. When "0", the time direction corresponding block and view direction corresponding block are not selected, but only neighboring blocks are selected. Can be.

In the motion vector prediction value MV _P , the motion vector prediction value MV _P is obtained and output using all or a part of the motion vectors of the block selected by the block selector 162. As described above, the motion vector prediction value MV _P may be obtained through an intermediate value calculation for all or part of the motion vectors of the selected block.

Motion vector prediction (MV _P Decoding device using the method to obtain

18 is a block diagram illustrating a configuration of an apparatus for decoding a multiview image according to an embodiment of the present invention. The decoding apparatus of FIG. 18 is a decoding apparatus of a multiview image using the method of obtaining the motion vector prediction value MV _P described above, and is a decoding apparatus corresponding to the encoding apparatus of FIG.

Referring to FIG. 18, the decoding apparatus includes a motion compensator 220, an entropy decoder 135, an inverse quantization and inverse changer 240, a motion vector predictor 250, and first and second adders 204. , 206). Hereinafter, the configuration and operation of the decoding apparatus will be described in detail with reference to FIG. 18.

The entropy decoder 235 performs an entropy decoding process on the input bitstream to generate quantization coefficients for the differential image of the current image to be reconstructed in block units. The inverse quantization and inverse transform unit 240 performs inverse quantization and inverse transformation on the quantization coefficients, reconstructs and outputs the difference image of the current image, and the difference image is input to the first adder 204.

Meanwhile, in the motion vector predicting unit 250, as in the encoding apparatus described with reference to FIGS. 15 and 17, the motion vector prediction value (MV) is obtained from the reference image of list 0, the reference image of list 1, and the information of the reconstructed current image. _P ) In determining the motion vector prediction value MV _P , whether to use the reference picture of List0 and / or the reference picture of List1 is a motion vector of a block corresponding to a time direction transmitted from an encoding apparatus or reproduced and input from a storage medium. (MV _T ) and / or the motion vector MV _V of the view direction correspondence block is determined based on the information in the slice unit. Since the detailed configuration and operation of the motion vector predictor 250 are the same as those of the motion vector predictor 150 of the encoding apparatus described above, a detailed description thereof will be omitted. That is, FIG. 17 and the related description may be equally applied to the motion vector predictor 250 according to the present embodiment.

The motion vector predictor MV _P generated by the motion vector predictor 250 is input to the second adder 206, and the second adder 250 receives the input motion vector difference MV _D and the motion vector predictor ( MV _P ) is added to generate and output a motion vector MV of the current block. The motion compensator 220 generates a predictive image by using the motion vector MV and information of the reference image of List0 and / or the reference image of List1. The first adder 204 reconstructs the current image by adding the predicted image input from the motion compensator 220 and the difference image input from the inverse quantization and inverse transform unit 240.

In the method of predicting a motion vector in multi-view video coding according to the present invention, a motion direction corresponding block of a temporal reference image in the same view as the view to which the current image belongs together with the motion vector MV _N of the neighboring block of the current block. A motion vector prediction value MV _P is obtained using the motion vector MV _V of the view direction corresponding block of the view direction reference image in the same time zone as the vector MV _T and / or the current image. Therefore, according to the present invention, a more accurate motion vector prediction value MV _P can be obtained in multi-view video coding, and as a result, the amount of information (entropy) of the motion vector difference value MV _D can be reduced to improve prediction coding efficiency. . That is, according to the present invention, it is possible to improve the predictive coding efficiency by effectively utilizing the similarity between images input from multiple cameras in multi-view video coding.

Claims (36)

A method for obtaining a motion vector prediction value in multi-view video coding, The reference image of the current block is the image of another view in the same time zone as the current image, And predicting the motion vector prediction value using one or more motion vectors adaptively selected from the motion vectors of the neighboring blocks of the current block and the motion vectors of the corresponding block in the time direction of the current block. The method of claim 1, Among the neighboring block and the time direction corresponding block And a motion vector prediction value using a motion vector of a block having a reference picture belonging to the same view as the view to which the reference picture of the current block belongs. The method of claim 1, Among the peripheral blocks, If there is a block having a reference picture belonging to the same view as the view to which the reference picture of the current block belongs, the motion vector of the block having a reference picture belonging to the same view as the view to which the reference picture of the current block belongs belongs is used. A motion vector prediction method comprising obtaining a motion vector prediction value. The method of claim 1, Out of the time-direction corresponding block And when there is no block having a reference image belonging to the same view as the view to which the reference image of the current block belongs, the motion vector prediction value is obtained using the motion vector of the neighboring block. A method for obtaining a motion vector prediction value in multi-view video coding, The reference image of the current block is an image of another time zone in the same view as the current image, And predicting the motion vector prediction value using one or more motion vectors adaptively selected from motion vectors of neighboring blocks of the current block and motion vectors of a view direction corresponding block of the current block. The method of claim 5, Among the peripheral block and the view direction corresponding block And a motion vector prediction value using a motion vector of a block having a reference image belonging to the same time zone that the reference image of the current block belongs to. The method of claim 5, Out of the surrounding blocks If there is a block having a reference image belonging to the same time zone that the reference image of the current block belongs to, the motion vector of the block having a reference image belonging to the same time zone as the reference time zone to which the reference image of the current block belongs belongs is used. A motion vector prediction method comprising obtaining a motion vector prediction value. The method of claim 5, Among the view direction corresponding block And if there is no block having a reference image belonging to the same time zone that the reference image of the current block belongs to, the motion vector prediction value is calculated using the motion vector of the neighboring block. A method for obtaining a motion vector prediction value in multi-view video coding, Of the motion vectors of the neighboring blocks of the current block, the motion vectors of the corresponding block in the time direction of the current block, and the motion vectors of the corresponding block in the view direction of the current block, a time zone in which the reference picture of the current block is the same as the current picture. And a motion vector prediction value using an adaptively selected motion vector, depending on whether the image is in another view or in another time zone in the same view as the current image. In the decoding method of a multiview image, When obtaining a motion vector prediction value, the motion vector of the neighboring block of the current block and the time direction correspondence of the current block are based on information indicating whether to use the motion vector of the block corresponding to the direction of time and the motion vector of the block corresponding to the direction of view. Obtaining the motion vector prediction value using at least one motion vector adaptively selected from a motion vector of a block and a motion vector of the view direction corresponding block of the current block; And And calculating a motion vector of the current block by adding the motion vector predicted value and a motion vector difference value of the input current block. The method of claim 10, Wherein the information is information included in a sequence header, information included in a slice header, or encoding information about a block. The method of claim 11, And the information is a variable length code, a fixed length code, or an Exp-Golomb code. The method of claim 10, wherein the obtaining of the motion vector prediction value When the information is information indicating that the motion vector of the corresponding block in the time direction and the motion vector of the corresponding block in the view direction are not used, only the motion vector of the neighboring block is used. If the information is information indicating that the motion vector of the corresponding block in the time direction and the motion vector of the corresponding block in the view direction are used, the motion vector of the neighboring block, the motion vector of the corresponding block in the time direction, and the view direction corresponding block are used. A decoding method of a multiview image, characterized by using at least one motion vector adaptively selected from among the motion vectors. In the decoding method of a multiview image, Based on information indicating whether to use a motion vector of a view-direction corresponding block when obtaining a motion vector prediction value, an adaptively selected from the motion vector of the neighboring block of the current block and the motion vector of the view-direction corresponding block of the current block Obtaining a motion vector prediction value of the current block using at least one motion vector; And And calculating a motion vector of the current block by adding the motion vector predicted value and a motion vector difference value of the input current block. 15. The method of claim 14, wherein obtaining the motion vector prediction value When the information indicates that the motion vector of the corresponding view direction block is not used, only the motion vector of the neighboring block is used. When the information is information indicating that the motion vector of the corresponding view direction block is used, one or more motion vectors adaptively selected from the motion vector of the neighboring block and the motion vector of the view direction corresponding block are used. Method of decoding a multiview image. In the multi-view video encoding method, Generating information indicating whether to use a motion vector of a time direction corresponding block and a motion vector of a view direction corresponding block when obtaining a motion vector prediction value; Based on the information, one or more motion vectors adaptively selected from among a motion vector of a neighboring block of the current block, a motion vector of a time direction corresponding block of the current block, and a motion vector of a view direction corresponding block of the current block, Obtaining a motion vector prediction value; And And subtracting the motion vector predicted value and the motion vector of the current block to obtain a motion vector difference value of the current block. In the multi-view video encoding method, Generating information indicating whether to use a motion vector of a view direction corresponding block when obtaining a motion vector prediction value; Obtaining a motion vector prediction value using at least one motion vector adaptively selected from a motion vector of a neighboring block of the current block and a motion vector of a view direction corresponding block of the current block based on the information; And And subtracting the motion vector predicted value and the motion vector of the current block to obtain a motion vector difference value of the current block. In the method of constructing the syntax of multi-view video coding, In the case of obtaining a motion vector prediction value, information indicating whether to use at least one of the motion vector of the corresponding block in the direction of time and the motion vector of the corresponding block in the view direction is included, wherein the motion of the corresponding block in the direction of time is based on the information. And a motion vector of at least one of a vector and a motion vector of the view direction corresponding block is used. In the decoding apparatus of a multiview image, The motion vector of the neighboring block of the current block, the time direction of the current block, based on input information indicating whether to use the motion vector of the block corresponding to the time direction and the motion vector of the block corresponding to the view direction when obtaining the motion vector prediction value. A motion vector predictor for obtaining the motion vector prediction value using a motion vector of at least one block adaptively selected from a motion vector of a corresponding block and a motion vector of the view direction corresponding block of the current block; And And an adder for adding the motion vector predicted value and the input motion vector difference to obtain a motion vector of the current block. In the decoding apparatus of a multiview image, Based on input information indicating whether to use a motion vector of a view-direction corresponding block when obtaining a motion vector prediction value, an adaptive value is selected from a motion vector of a neighboring block of a current block and a motion vector of the view-direction corresponding block of the current block. A motion vector predictor for obtaining the motion vector prediction value by using motion vectors of at least one block selected as And And an adder for adding the motion vector predicted value and the input motion vector difference to obtain a motion vector of the current block. In the multi-view video encoding apparatus, A usage determination unit for generating information indicating whether to use a motion vector of a time direction corresponding block and a motion vector of a view direction corresponding block when obtaining a motion vector prediction value; Based on the information, the motion vector of at least one block adaptively selected from among the motion vector of the neighboring block of the current block, the motion vector of the corresponding block in the time direction of the current block, and the motion vector of the corresponding block in the view direction of the current block. A motion vector predictor for obtaining the motion vector predicted value using the motion vector predictor; And And an adder for subtracting the motion vector predicted value and the motion vector of the current block to obtain a motion vector difference value of the current block. In the multi-view video encoding apparatus, A usage determination unit for generating information indicating whether to use a motion vector of a view direction corresponding block when obtaining a motion vector prediction value; A motion vector predictor for obtaining the motion vector prediction value using at least one motion vector adaptively selected from a motion vector of a neighboring block of the current block and a motion vector of a view direction corresponding block of the current block based on the information; And And an adder for subtracting the motion vector predicted value and the motion vector of the current block to obtain a motion vector difference value of the current block. delete delete delete delete delete delete delete delete delete delete delete delete delete delete

Images