JP2002010269A - Motion vector detection method and video encoding device - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To predict a motion vector that does not match the motion of an image, to detect a motion vector having a smaller amount of encoded information, and to encode the motion vector. An information amount of information is suppressed to improve the coding efficiency of a moving image coding device. SOLUTION: In a motion vector detecting method for dividing a moving image into predetermined blocks and detecting a motion vector of the moving image for each block, a plurality of motion vector candidates are stored in a storage unit based on a prediction error evaluation reference value. Comparing the prediction error evaluation reference values of the stored plurality of motion vector candidates with each other, and based on a comparison value difference signal obtained by the comparison, any one of the plurality of motion vector candidates is extracted from the plurality of motion vector candidates. Select and detect as a motion vector.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting method for dividing a moving image into small blocks and detecting a motion vector of the moving image for each block in a system for highly efficient coding of a moving image and digital transmission. , A video encoding device.

[0002]

2. Description of the Related Art Conventionally, as a technique for compressing, encoding and transmitting a moving image, there is a technique using motion compensation prediction. This technology divides a frame image constituting a moving image into blocks, and creates an encoded image from a future or past reference image for each block. In order to perform the motion compensation prediction, it is necessary to detect a motion vector. One of the motion vector detection methods is MPEG-2 (Moving), which is an international standard for moving picture coding.
Picture Coding ExpertGro
up) test model, ISO / IEC JTC1 / S
C29 / WG11 / N0400 "TEST MODE"
L5 ".

This method is basically a method called a block matching method. For example, a current image is divided into blocks (for example, 16 pixels × 16 lines or 16 pixels × 16 pixels).
8 lines), a prediction error evaluation criterion value is calculated for the block using an error evaluation criterion such as a sum of absolute difference values, and a block having the smallest prediction error evaluation criterion value is obtained from the reference image. . Then, a method of detecting a position shift occurring between the blocks as a motion vector of the block. In MPEG-2, 1
A “block” consisting of 6 pixels × 16 lines is particularly referred to as a “macro block”.

When bidirectional prediction is performed, prediction from a temporally past image (forward prediction) and prediction from a temporally future image (reverse prediction) are performed by block matching methods. After that, the prediction error evaluation reference value is obtained using a reference image created by performing one-to-one linear interpolation at the position of each selected motion vector candidate. Then, of the motion vector candidates searched for in the three prediction directions, the motion vector candidate with the smallest prediction error evaluation reference value is detected as the motion vector.

[0005]

However, in the above-described motion vector detection method, the prediction error evaluation reference value for the block on the current image is calculated using the error evaluation reference value such as the sum of absolute differences. The smallest block is obtained from the reference image. Therefore, a motion vector matching the motion of the image is not always detected, and a motion vector not matching the motion of the image may be detected. This tendency was particularly remarkable when the image contained noise. If a motion vector that does not match the motion of the image is detected, a large amount of information is allocated to the encoded information of the unnecessary motion vector,
There has been a problem that the coding efficiency of the moving picture coding apparatus is reduced.

The present invention has been made to solve such a problem. When it is predicted that a motion vector not matching the motion of an image is detected, a motion vector having a smaller amount of encoded information is used. It is an object of the present invention to improve the coding efficiency of a moving image coding apparatus by preventing a large amount of information from being divided into the amount of information of coding information of a motion vector.

[0007]

SUMMARY OF THE INVENTION The present invention is a motion vector detecting method for dividing a moving image into predetermined blocks, and detecting a motion vector of the moving image for each block. A plurality of motion vector candidates are stored in storage means, and the prediction error evaluation reference values of the stored plurality of motion vector candidates are compared with each other. Based on a comparison value difference signal obtained by the comparison, the plurality of In this case, one of the motion vector candidates is selected from the motion vector candidates, and the selected motion vector candidate is used as a motion vector.

In the present invention, the comparison value difference signal is a signal indicating a degree of dispersion of the prediction error evaluation reference value or a signal indicating a difference between the prediction error evaluation reference values.

Further, the present invention provides a motion vector candidate having the smallest prediction error evaluation reference value, a motion vector candidate having the minimum number of vectors added to one block, and a motion vector candidate having a minimum length. A motion vector candidate is selected and set as a motion vector.

Further, in the present invention, the plurality of motion vector candidates stored in the storage means include at least one motion vector candidate searched for each of a plurality of prediction directions.

Further, the present invention provides a moving picture coding system which divides a moving picture into predetermined blocks, detects a motion vector of the moving picture for each block, performs motion compensation prediction, and codes the moving picture. A storage means for storing a plurality of motion vector candidates; a comparison means for comparing prediction error evaluation reference values of the plurality of motion vector candidates stored in the storage means; and a comparison value obtained by the comparison. Detecting means for selecting a motion vector candidate for each block of the moving image from the plurality of motion vector candidates based on the difference signal, and using the selected motion vector as a motion vector for each block of the moving image.

In the present invention, the comparison value difference signal is a signal indicating a degree of dispersion of the prediction error evaluation reference value or a signal indicating a difference between the prediction error evaluation reference values.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 An embodiment of a moving picture coding apparatus for performing motion compensation prediction using a motion vector according to the present invention will be described with reference to FIG. Figure 1
FIG. 1 is a functional block diagram showing a configuration of a video encoding device according to Embodiment 1 of the present invention. In FIG. 1, reference numeral 1 denotes an input image signal which is an input image signal, 2 denotes a rearranging unit for rearranging the frame order of an input image obtained from the input image signal 1, and 3 denotes a past connected to the output of the rearranging unit 2 Alternatively, the subtractor 4 for taking the difference between the future image and the input image is connected to the output side of the subtractor 3 and, for example, DCT (Disc)
An orthogonal transform unit that performs orthogonal transform using “rete Cosine Transform (discrete cosine transform)”. Reference numeral 5 denotes a quantization unit connected to the output side of the orthogonal transformation unit 4 and performs quantization for information compression. Reference numeral 6 denotes a quantization unit connected to the output side of the quantization unit 5 and encodes a signal according to the frequency of data generation. The variable length coding unit 7 to be assigned is a buffer memory that temporarily stores data output from the variable length coding unit 6 and then outputs the data.

Reference numeral 8 denotes coded data output from the buffer memory 7, 9 denotes an inverse quantization unit connected to the output side of the quantization unit 5 and performs inverse quantization of the quantized data.
Is an inverse orthogonal transform unit connected to the output side of the inverse quantization unit 9 and performs inverse orthogonal transform on the orthogonally transformed data, 11 is an adder connected to the output side of the inverse orthogonal transform unit 10, and 12 is an adder 1
Reference numeral 13 denotes an image memory connected to the output side of the adder 1 and stores a past reference image. Reference numeral 13 denotes an image memory connected to the output side of the adder 11 and stores a future reference image.

Reference numeral 14 denotes a motion compensating unit which is connected to the output side of the image memories 12 and 13 and performs motion compensation prediction according to a past or future reference image and a detected motion vector. Reference numeral 15 denotes a motion vector detecting unit which is connected to the output side of the image memories 12 and 13 and detects a motion vector from the input image and the reference image, and detects the motion vector detected by the variable-length coding unit 6 and the motion compensation unit 14. And other data.

Next, the operation of the moving picture coding apparatus for performing motion compensation prediction using the motion vector configured as described above will be described. First, the input image signal 1 is rearranged in the frame order of the input image by the rearranging unit 2, and the order of frames that are temporally earlier or later is rearranged. In order to support the bi-directional (bi-directional) prediction employed in MPEG-2, which is an international standard for video coding, it is necessary to code a reference image that is temporally future ahead of the coded image. Therefore, the frame order is rearranged.

Next, the difference between the input image signal 1 and the motion-compensated prediction signal output from the motion compensator 14 is calculated by the subtractor 3. Then, for example, DCT (Disc)
Orthogonal transform is performed using a "rete Cosine Transform (discrete cosine transform)", and the redundancy in the spatial direction is reduced. After that, the quantizing unit 5 performs quantization to reduce the code amount, and the variable length coding unit 6 assigns a code using the appearance probability of the code. The encoded data 8 assigned by the variable length encoding unit 6 is output after being temporarily stored in a buffer memory 7. When intra coding (intra-frame coding) is selected, zero coding is sent to the subtractor 3 and the adder 11 instead of the output of the motion compensator 14, so that intra coding is performed. .

Here, the generation of the motion compensation prediction signal output from the motion compensation section 14 is performed as follows. The output of the quantization unit 5 is inversely quantized by an inverse quantization unit 9, subsequently inversely orthogonally transformed by an inverse orthogonal transformation unit 10, and further added to a motion-compensated prediction image signal by an adder 11 to form an image memory. 12 or the image memory 13. Here, there are two image memories,
This is because a temporally past image is accumulated on one side and a temporally future image is accumulated on the other side, and both forward and backward (bidirectional) prediction is performed. That is, the prediction of the current image is
The prediction is performed in three variations: forward prediction from a past image stored in the image memory 12, backward prediction from a future image stored in the image memory 13, and bidirectional (bidirectional) prediction.

The motion vector detecting section 15 uses the image signal of the past or future reference image output from the image memory 12 and the image memory 13 and the image signal of the current image output from the rearranging section 2. A motion vector is detected. The motion compensation unit 14 receives the data of the motion vector detected by the motion vector detection unit 15 and
The block indicated by the corresponding motion vector is output to the subtractor 3 by using the reference image output from each of the reference images 2 and 13.

A method of detecting a motion vector by the motion vector detecting section 15 will be described below with reference to FIG. FIG. 2 is a flowchart of motion vector detection of the moving picture coding apparatus according to Embodiment 1 of the present invention. First, the motion vector detecting unit 15 searches for a motion vector candidate which is a candidate for a motion vector finally detected within a predetermined search range (step 101). Hereinafter, “step” is abbreviated as “S”. At this time, the motion vector detecting unit 15 determines not only the motion vector candidates having the smallest prediction error evaluation reference value in the searched range but also the α (α is a natural number) The motion vector candidate and the prediction error evaluation reference value corresponding to the motion vector candidate are both stored in the storage means (S102).

Then, a variance value (a signal indicating the degree of variance) as a comparison value difference signal obtained by comparing the prediction error evaluation reference value with the α motion vector candidates stored in the storage means is calculated. The obtained dispersion value is set to β (S1
03). The obtained variance β is compared with a predetermined threshold TH (S104), and the following processing is performed based on the result.

If the variance β is equal to or larger than the predetermined threshold TH (NO in S104), the prediction of the motion vector candidate having the smallest prediction error evaluation reference value among the α motion vector candidates stored is performed. It is determined that the difference between the error evaluation reference value and the prediction error evaluation reference values of other motion vector candidates is large. Since the motion vector candidate with the smallest prediction error evaluation reference value matches the motion of the image, the prediction error evaluation reference value may be smaller than the prediction error evaluation reference values of other motion vector candidates. Is high. Therefore, in this case, a motion vector candidate having the smallest prediction error evaluation reference value is selected from among the stored α motion vector candidates, and detected as a motion vector (S105).

If the variance β is smaller than the predetermined threshold TH (YES in S104), the prediction error evaluation reference values of α motion vector candidates are not dispersed, and the difference between the prediction error evaluation reference values is different. There is not much. Also, the motion vector candidate having the smallest prediction error evaluation reference value is likely to not match the motion of the image due to the influence of noise or the like. Therefore, in this case, the motion vector detecting unit 15 uses the stored α motion vector candidates as one block at the time of encoding,
For example, a motion vector candidate searched for from a combination of a prediction direction and a prediction mode in which the number of motion vector candidates added to one macroblock of 16 pixels × 16 lines is the smallest (S106).

For example, in MPEG-2, which is an international standard for moving picture coding, forward prediction (forward prediction) for performing prediction from a temporally past image, and prediction after performing temporal prediction from a future image. There are variations such as directional prediction (backward prediction) and bidirectional prediction (bidirectional prediction) that performs prediction from both temporally past images and future images, and each of them has a prediction mode such as frame prediction or field prediction. Exists. If the combination of the prediction direction and a prediction mode other than the prediction direction, which is a prediction method, is a prediction type, the number of motion vector candidates added to one block (for example, a macroblock) at the time of encoding differs depending on the prediction type.

For example, in MPEG-2, the number of motion vector candidates added to one block is doubled in the case of bidirectional prediction as opposed to unidirectional prediction such as forward prediction and backward prediction. On the other hand, also in field prediction, the number of motion vector candidates added to one block with respect to frame prediction is doubled. Then, at the time of encoding, a plurality of motion vectors added to one block are set as one set,
A motion vector may be detected. The smaller the number of motion vectors in a set of motion vectors, the more the amount of encoded motion vector information can be reduced. Therefore,
As described above, a motion vector candidate searched for by the prediction type with the least number of vectors added to one block is selected and detected as a motion vector.

However, when there are a plurality of motion vector candidates having the minimum number of motion vector candidates added to one block (S
107 (YES), the motion vector detection unit 15 selects a motion vector candidate with the minimum vector length among them, and detects it as a motion vector (S108). In addition,
In the case where the number of motion vector candidates added to one block is two or more, the sum of the lengths of the number of vectors is obtained even when the ones having the longest vector length added to one block are compared. The motion vector candidates with the minimum vector length may be selected using any method of comparing the vector lengths with each other.

When there is no plurality of motion vector candidates with the minimum number of motion vector candidates added to one block (S1
(NO in 07), the motion vector detection unit 15 detects the motion vector candidate selected in S106 as a motion vector.

In the above-described configuration and operation, if it is highly probable that a motion vector that matches the motion of the image is not selected due to, for example, noise, it is added to one block of a plurality of stored motion vector candidates. The motion vector candidates searched for in the prediction type that minimizes the number of motion vector candidates to be performed are selected. Therefore, it is possible to prevent the information amount from being excessively divided by the information amount (the number of bits) of the encoded information of the motion vector predicted not to match the motion of the image.
It is possible to suppress the information amount of the encoded information of the motion vector with respect to the entire encoded information amount, and to improve the encoding efficiency.

Since the candidate having the minimum length of the motion vector candidate is selected, the information amount of the encoded information of the motion vector can be further suppressed, and the encoding efficiency can be improved.

Further, by suppressing the information amount of the coded information of the motion vector with respect to the entire coded information amount, the information amount allocated to the coded information of the motion compensation prediction error signal is secured, and The prediction error signal is not very coarsely quantized, and the image quality can be improved.

In the present embodiment, the stored α
Although the variance of the prediction error evaluation reference values for the motion vector candidates is obtained, any method other than the variance may be used as long as the method can determine the degree of dispersion of a plurality of prediction error evaluation reference values. A method is also acceptable. For example, a method using the sum of absolute differences may be used, and an appropriate method can be selected in relation to the amount of calculation for calculating the degree of variation.

Embodiment 2 FIG. The basic configuration of the moving picture coding apparatus according to the present embodiment is the same as that of the first embodiment, and a description of the components will be omitted. Hereinafter, a method of detecting a motion vector by the motion vector detection unit 15 according to the present embodiment will be described with reference to FIG. FIG. 3 is a flowchart of motion vector detection of the moving picture coding apparatus according to Embodiment 2 of the present invention.

In FIG. 3, first, the motion vector detecting section 15 searches for a motion vector candidate which is a candidate for a motion vector finally detected in a predetermined search range (S201). At this time, the motion vector detection unit 15
Is a motion vector candidate having the smallest prediction error evaluation reference value within the searched range, as well as α (α is a natural number) motion vector candidates in descending order of the prediction error evaluation reference value, and the corresponding motion vector candidates. The prediction error evaluation reference value is stored in the storage means together (S202).

The motion vector detecting section 15 sets A as the prediction error evaluation reference value of the motion vector candidate having the minimum prediction error evaluation reference value among the α motion vector candidates stored in the storage means. Also, the prediction error evaluation reference value of the motion vector candidate searched for by the prediction type that minimizes the number of motion vector candidates added to one block is set to B (step S203). Here, when there are a plurality of motion vector candidates with the minimum number of motion vector candidates added to one block, the lengths of the motion vector candidates are compared with each other to predict the motion vector candidate having the minimum length. Let B be the error evaluation reference value.

The absolute value of (AB), which is a signal indicating the difference between A and B calculated as described above, is compared with a preset threshold value C (S204). Is performed. If the absolute value of (AB) is equal to or larger than the preset threshold C (NO in S204), it is determined that the difference between the prediction error evaluation reference value A and the prediction error evaluation reference values of other motion vector candidates is large. I do. Since the motion vector candidate with the smallest prediction error evaluation reference value matches the motion of the image, the prediction error evaluation reference value may be smaller than the prediction error evaluation reference values of other motion vector candidates. Is high. Therefore, the motion vector detection unit 15 selects a motion vector candidate with the smallest prediction error evaluation reference value and detects it as a motion vector (step S20).
5).

If the absolute value of (AB) is smaller than a predetermined constant value C, there is a high possibility that there is not much difference between the prediction error evaluation reference values of the α motion vector candidates. Also, the motion vector candidate having the smallest prediction error evaluation reference value is likely to not match the motion of the image due to the influence of noise or the like. Therefore, the motion vector detection unit 15
Selects, from among the stored α motion vector candidates, a motion vector candidate searched for in a prediction type that minimizes the number of motion vector candidates added to one block during encoding (S206).

However, when there are a plurality of motion vector candidates having the minimum number of motion vector candidates added to one block (S
(YES in 207), the motion vector detection unit 15 selects a motion vector candidate having the minimum vector length among them, and detects it as a motion vector (S208). In addition,
In the case where the number of motion vector candidates added to one block is two or more, the sum of the lengths of the number of vectors is obtained even when the ones having the longest vector length added to one block are compared. The motion vector candidates with the minimum vector length may be selected using any method of comparing vector lengths, or any other method of comparing vector lengths.

When there is no plurality of motion vector candidates with the minimum number of motion vector candidates added to one block (S2
(NO in 07), the motion vector detection unit 15 detects the motion vector candidate selected in S206 as a motion vector.

In the above configuration and operation, it is possible to predict whether or not the motion vector matches the motion of the image by relatively simple processing. For example, the number of motion vector candidates stored in the storage means is enormous. In this case, it is useful in terms of processing time. Then, by preventing the information amount from being excessively divided by the information amount (the number of bits) of the coded information of the motion vector predicted not to match the motion of the image, the coding of the motion vector with respect to the entire coded information amount is prevented. It is possible to suppress the information amount of the coded information and improve the coding efficiency.

Embodiment 3 The basic configuration of the moving picture coding apparatus according to the present embodiment is the same as that of the first embodiment, and a description of the components will be omitted. Hereinafter, a method of detecting a motion vector by the motion vector detection unit 15 according to the present embodiment will be described with reference to FIG. FIG. 4 is a flowchart of motion vector detection of the video encoding device according to Embodiment 3 of the present invention.

In FIG. 4, first, the motion vector detecting section 15 searches for a motion vector candidate which is a candidate for a motion vector finally detected within a predetermined search range (S301). At this time, the motion vector detection unit 15
Are α motion vectors (α is a natural number) in order from the motion vector with the smallest prediction error evaluation criterion value in the searched range, The error evaluation reference value is stored together in the storage means (S302).

Next, apart from the α motion vector candidates stored in S302, a motion vector candidate having the minimum error evaluation reference value for each prediction type described in the second embodiment and a corresponding error evaluation The reference value is stored in the storage unit (Step S303).

Then, a variance value (a signal indicating the degree of variance), which is a comparison value difference signal obtained by comparing the prediction error evaluation reference values for the α motion vector candidates stored in S302, is obtained. The value is set to β (S30
4). The obtained variance β and a predetermined threshold TH
(S305), and the following processing is performed based on the result.

If the variance β is equal to or greater than the predetermined threshold TH (NO in S305), a motion vector candidate with the smallest prediction error evaluation reference value is selected from the stored α motion vector candidates. Detect as a motion vector (S3
06).

If the variance β is smaller than the predetermined threshold TH (YES in S305), the α stored in S302
Among the motion vector candidates, the prediction error evaluation reference value of the motion vector candidate having the smallest prediction error evaluation reference value is A (S307). Then, the absolute value of the difference between A and the prediction error evaluation reference value of the α motion vectors stored in S302 and the motion vector stored for each prediction type in S303 is calculated. Is deleted from the storage (S308). If the difference is larger than the threshold value D, it is more likely that the difference does not match the motion of the image.

Next, the motion vector detecting section 15 executes S30
The motion vector candidates searched for by the prediction type in which the number of motion vector candidates added to one block, for example, one macroblock of 16 pixels × 16 lines at the time of encoding, from the motion vector candidates not deleted in step 8, is minimized. Is selected (S309).

However, when there are a plurality of motion vector candidates having the minimum number of motion vector candidates added to one block (S
310 (YES), the motion vector detection unit 15 selects a motion vector candidate having the minimum vector length among them, and detects it as a motion vector (S311). In addition,
In the case where the number of motion vector candidates to be added to one block is two or more, the sum of the lengths of the number of vectors is the same even when the longest vector added to one block is compared. The motion vector candidates with the minimum vector length may be selected using any method of comparing the vector lengths with each other.

When there is no plurality of motion vector candidates with the minimum number of motion vector candidates added to one block (S3
(NO in 10), the motion vector detection unit 15 detects the motion vector candidate selected in S309 as a motion vector.

In the above configuration and operation, at least one motion vector candidate is stored for each prediction type. Therefore, even when the stored α motion vector candidates are concentrated on the same prediction type, all the motion vector candidates are stored. A motion vector can be detected from the motion vector candidates searched for by type. Further, since the number of motion vector candidates is increased as compared with the first embodiment, a motion vector candidate having a smaller amount of information (number of bits) of encoded information of the motion vector can be detected as a motion vector. Encoding efficiency can be improved.

Embodiment 4 The basic configuration of the moving picture coding apparatus according to the present embodiment is the same as that of the first embodiment, and a description of the components will be omitted. Hereinafter, a method of detecting a motion vector by the motion vector detection unit 15 according to the present embodiment will be described with reference to FIG. FIG. 5 is a flowchart of motion vector detection of the moving picture coding apparatus according to Embodiment 4 of the present invention.

Referring to FIG. 5, first, the motion vector detecting section 15 searches for a motion vector candidate which is a candidate for a motion vector finally detected within a predetermined search range (S401). At this time, the motion vector detection unit 15
Are α motion vectors (α is a natural number) in order from the motion vector with the smallest prediction error evaluation criterion value in the searched range, The error evaluation reference value is stored in the storage means together (S402). In addition, separately from the α motion vector candidates stored in S402, the motion vector candidate having the smallest error evaluation reference value for each prediction type and the corresponding error evaluation reference value are stored in the storage unit (S40).
3).

Next, the motion vector detecting section 15 sets the prediction error evaluation reference value of the smallest motion vector candidate of the prediction error evaluation reference value to A out of the stored α motion vector candidates.
And The prediction error evaluation reference value of the motion vector candidate searched for by the prediction type that minimizes the number of motion vector candidates added to one block is set to B (S4
04). Here, when there are a plurality of motion vector candidates to be added to one block in which the number of motion vector candidates is the minimum, the lengths of the motion vector candidates are compared with each other to predict the motion vector candidate having the minimum length. The error evaluation reference value is set to B (S404).

The absolute value of (AB), which is a signal indicating the difference between A and B calculated as described above, is compared with a preset threshold value C (S405). Is performed. If the absolute value of (AB) is equal to or greater than the preset threshold C (NO in S405), the motion vector detection unit 15 selects a motion vector candidate with the smallest prediction error evaluation reference value, and sets the motion vector candidate as the motion vector. It is detected (step S406).

On the other hand, if the absolute value of (AB) is smaller than the preset threshold value C (YES in S405), S40
The absolute value of the difference between A and the prediction error evaluation reference value of the α motion vectors stored in step S2 and the motion vector stored in step S403 for each prediction type is calculated.
Those larger than are deleted from the storage (S407).

Then, the motion vector detecting unit 15 determines in S4
From the motion vector candidates that have not been deleted in 07, a motion vector candidate searched for by a prediction type that minimizes the number of motion vector candidates added to one block during encoding is selected (S408).

However, when there are a plurality of motion vector candidates having the minimum number of motion vector candidates added to one block (S
(YES in S409), the motion vector detection unit 15 selects a motion vector candidate having the minimum vector length among them, and detects it as a motion vector (S410). In addition,
In the case where the number of motion vector candidates to be added to one block is two or more, the sum of the lengths of the number of vectors is the same even when the longest vector added to one block is compared. The motion vector candidates with the minimum vector length may be selected using any method of comparing vector lengths, or any other method of comparing vector lengths.

If there are no plural motion vector candidates with the minimum number of motion vector candidates added to one block (S4
09 (NO), the motion vector detection unit 15 detects the motion vector candidate selected in S408 as a motion vector.

In the above configuration and operation, whether or not the motion vector matches the motion of the image can be predicted by relatively simple processing. For example, when the number of stored motion vector candidates is enormous, Useful in terms of processing time. Then, by preventing the information amount from being excessively divided by the information amount (the number of bits) of the coded information of the motion vector predicted not to match the motion of the image, the coding of the motion vector with respect to the entire coded information amount is prevented. It is possible to suppress the information amount of the coded information and improve the coding efficiency.

[0059]

As is apparent from the above invention, the motion vector detecting method according to the present invention provides a method for estimating the amount of information (the number of bits) of encoded information of a motion vector predicted not to match the motion of an image. Preventing the amount of information from being over-
It is possible to suppress the information amount of the encoded information of the motion vector with respect to the entire encoded information amount, and to improve the encoding efficiency.

Further, the moving picture coding apparatus according to the present invention is capable of dividing the information amount excessively into the information amount (the number of bits) of the coding information of the motion vector predicted not to match the motion of the picture. Can be prevented, the amount of motion vector coded information with respect to the entire coded information amount can be suppressed, and coding efficiency can be improved.

[Brief description of the drawings]

FIG. 1 is a functional block diagram illustrating a configuration of a video encoding device according to Embodiment 1 of the present invention.

FIG. 2 is a diagram showing a flow of motion vector detection of the video encoding device according to the first embodiment of the present invention.

FIG. 3 is a diagram showing a flow of motion vector detection of the moving picture coding apparatus according to Embodiment 2 of the present invention.

FIG. 4 is a diagram showing a flow of motion vector detection of a video encoding device according to Embodiment 3 of the present invention.

FIG. 5 is a diagram showing a flow of motion vector detection of the moving picture coding apparatus according to Embodiment 4 of the present invention.

[Explanation of symbols]

1 input image signal, 2 reordering unit, 3 subtractor,
4 orthogonal transform unit, 5 quantizer, 6 variable length encoder, 7 buffer, 8 encoded data, 9 inverse quantizer, 10 inverse orthogonal transform unit, 11 adder, 1
2,13 Image memory, 14 Motion compensation unit, 15
Motion vector detection unit.

 ────────────────────────────────────────────────── ─── Continued on the front page F term (reference) 5C059 KK19 LA09 MA00 MA04 MA05 MA14 MA23 MC11 MC38 ME01 NN01 NN21 NN28 PP04 TA62 TB07 TC12 TD02 TD04 TD06 TD12 UA02 UA33

Claims (6)

[Claims] 1. A motion vector detecting method for dividing a moving image into predetermined blocks and detecting a motion vector of the moving image for each block, wherein a plurality of motion vector candidates are stored based on a prediction error evaluation reference value. And remember
The stored prediction error evaluation reference values of the plurality of motion vector candidates are compared with each other, and any one of the plurality of motion vector candidates is selected from the plurality of motion vector candidates based on a comparison value difference signal obtained by the comparison. And a motion vector detecting method. 2. The method according to claim 1, wherein the comparison value difference signal is a signal indicating a degree of dispersion of the prediction error evaluation reference value or a signal indicating a difference between the prediction error evaluation reference values. Motion vector detection method. 3. A motion vector candidate having the smallest prediction error evaluation reference value, a motion vector candidate having the smallest number of vectors added to one block, and a motion vector candidate having a minimum length. 3. The motion vector detecting method according to claim 1, wherein the selected motion vector is selected as a motion vector. 4. The method according to claim 1, wherein the plurality of motion vector candidates stored in the storage unit include at least one motion vector candidate searched for each of a plurality of prediction directions. 3. The motion vector detection method according to 3. 5. A moving picture coding apparatus for dividing a moving picture into predetermined blocks, detecting a motion vector of the moving picture for each block, performing motion compensation prediction, and coding the moving picture. Storage means for storing a plurality of motion vector candidates, comparison means for comparing prediction error evaluation reference values of the plurality of motion vector candidates stored in the storage means, and a comparison value difference signal obtained by the comparison. Detecting means for selecting a motion vector candidate for each block of the moving image from the plurality of motion vector candidates and using the selected motion vector as a motion vector for each block of the moving image. apparatus. 6. The signal according to claim 5, wherein the comparison value difference signal is a signal indicating a degree of dispersion of the prediction error evaluation reference value or a signal indicating a difference between the prediction error evaluation reference values. Video encoding device.