KR100664322B1 - Image processing apparatus capable of noise reduction and image processing method for noise reduction - Google Patents

Abstract

An apparatus for removing noise of a video signal and a method for removing noise thereof are disclosed. The first noise removing unit obtains the first minimum motion prediction error value and the motion estimation information of the current block through the motion prediction method, removes the noise of the motion estimation information, and outputs the first candidate image, and the second noise removing unit detects the motion. The second minimum motion prediction error value and the motion detection information are obtained through the method, and the second candidate image is output by removing the noise of the motion detection information, and the comparison determiner determines the number and the predetermined number of the first and second minimum motion prediction error values. A reference flag signal for selecting an optimal image is determined by comparing the thresholds of the optimal values, and the optimum image selecting unit selects and outputs one of the first and second candidate images as the optimal image based on the reference flag signal. Accordingly, the natural image may be output by determining the optimal image by comparing the number of motion prediction error values calculated in different ways.

Description

Image processing apparatus having function of noise reduction and image processing method for noise reduction thereof

1 is a block diagram showing a conventional noise removing device;

FIG. 2 is a diagram for explaining an unnatural image generated by FIG. 1;

3 is a block diagram schematically illustrating an apparatus for removing noise of a video signal according to an exemplary embodiment of the present invention;

4 is a block diagram illustrating in detail the comparison determiner of FIG. 3;

FIG. 5 is a diagram for describing a third SAD value of a line unit stored in the memory unit of FIG. 4; FIG.

FIG. 6 is a flowchart for schematically describing a method for removing noise of an image signal implemented by FIG. 3.

Description of the main parts of the drawing

300: noise removing device 310: first noise removing unit

320: second noise removing unit 330: comparison determination unit

332: comparison unit 334: memory unit

336: first counter 338: second counter

339: Determining unit 340: Optimal image selecting unit

The present invention relates to an image processing apparatus capable of noise reduction and an image processing method for noise reduction thereof. More particularly, a natural image is obtained by determining an optimal image by comparing the number of motion prediction error values calculated in different ways. An image processing apparatus capable of outputting noise reduction and an image processing method for noise reduction thereof.

In general, noise is mixed in a video signal applied to a video signal processing apparatus such as a television or a video tape recorder, and noise mixed in the video signal causes a great deterioration in image quality. Therefore, various techniques for removing noise of an image signal have been developed, and the technique serves as an important factor in determining an image quality.

1 is a block diagram illustrating a conventional image processing apparatus capable of removing noise.

Referring to FIG. 1, the conventional first noise removing unit 10 calculates motion estimation information for each block of a current frame by comparing with a previous frame, and then removes noise by filtering. (info_1) is output and a first minimum SAD (Sum of Absolute Difference) value of each block is calculated.

The second noise removing unit 20 calculates motion detection information for each block of the current frame by comparing the previous frame with the previous frame, and outputs motion detection information info_2 from which noise is removed through filtering. Compute a second minimum SAD value of the block.

The comparison selecting unit 30 compares the inputted first minimum SAD value with the second minimum SAD value to correspond to the smaller SAD value, that is, the motion estimation information info_1 from which the noise is removed and the motion from which the noise is removed. One of the detection information info_2 is selected and output as an optimum video signal.

However, when the optimal video signal is determined by comparing the first minimum SAD value and the second minimum SAD value as described above, an unnatural image is displayed as illustrated in FIG. 2.

Referring to FIG. 2, most of the entire image has a result of removing noise by using motion estimation information (indicated by a dashed line), and only three areas of a result of removing noise by using motion detection information (indicated by black). Have In this case, the entire image is displayed as an unnatural image that can be easily recognized by the viewer as shown in FIG. 2 because the boundary of each region is inconsistent due to the peripheral influence factor that is involved in the calculation result of the two results, that is, the gain. do. That is, conventionally selecting an optimal video signal by comparing only the first minimum SAD value and the second minimum SAD value has a limitation in improving image quality.

An object of the present invention is to prevent unnatural images generated by simply selecting an optimal image based on a smaller value among a first minimum SAD value based on motion prediction and a second minimum SAD value based on motion detection. The present invention provides an image processing apparatus capable of noise reduction, and an image processing method for noise reduction thereof.

In order to solve the above technical problem, the image processing apparatus capable of noise reduction according to the present invention compares a current search block of a current frame with a predetermined search region of a previous frame, and compares a first minimum motion prediction error value and a motion of the current block. A first noise removing unit for obtaining estimation information and outputting a first candidate image by removing noise of the motion estimation information; Comparing a previous block corresponding to the same position as the current block among the previous frames, a second minimum motion prediction error value and motion detection information of the current block are obtained, and a second candidate image is removed by removing noise of the motion detection information. An output second noise removing unit; A comparison determination unit determining a reference flag signal for selecting an optimal image by comparing the output first and second minimum motion prediction error values with a predetermined threshold value; And an optimal image selector configured to selectively output one of the first and second candidate images as the optimal image based on the determined reference flag signal.

More specifically, the comparison determiner, the comparison unit for outputting a smaller value of the first and second minimum motion prediction error value; A memory unit which stores the output small value in line units; A counter for counting each of the first and second minimum motion prediction error values included in any window of a predetermined number of lines stored in the memory unit; And a determiner configured to determine the reference flag signal based on the counted result and the predetermined threshold value.

The determining unit may output a reference flag signal corresponding to any one of the counted first and second minimum motion prediction error values to the optimal image selector if the number of the counted first and second minimum motion prediction error values is greater than or equal to the threshold value. The image selector selects a candidate image corresponding to the output reference flag signal among the first and second candidate images as the optimal image.

On the other hand, to solve the above technical problem, the image processing method for noise reduction according to the present invention, (a) the first minimum motion of the current block by comparing the current block of the current frame and the predetermined search region of the previous frame Obtaining a prediction error value and motion estimation information, and outputting a first candidate image by removing noise of the motion estimation information; (b) comparing a previous block corresponding to the same position as the current block in the previous frame to obtain a second minimum motion prediction error value and motion detection information of the current block, and removing noise of the motion detection information to remove the second block; Outputting a candidate image; (c) determining a reference flag signal for selecting an optimal image by comparing a predetermined threshold value of the number of output first and second minimum motion prediction error values; And (d) selectively outputting one of the first and second candidate images as the optimal image based on the determined reference flag signal.

More specifically, the step (c) may include: (c1) outputting a smaller value of the first and second minimum motion prediction error values; (c2) storing the output small value in units of lines; (c3) counting each of the first and second minimum motion prediction error values included in any window of the stored predetermined number of lines; And (c4) outputting a reference flag signal corresponding to any one of the counted first and second minimum motion prediction error values equal to or greater than the predetermined threshold value. In step d), the candidate image corresponding to the flag signal output in step (c4) among the first and second candidate images is selected as the optimal image.

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

3 is a block diagram schematically showing an image processing apparatus capable of attenuating noise according to an exemplary embodiment of the present invention, and FIG. 4 is a block diagram illustrating in detail the comparison determiner of FIG. 3.

Referring to FIG. 3, an image processing apparatus 300 capable of attenuating noise according to an exemplary embodiment of the present invention may include a first noise remover 310, a second noise remover 320, a comparison determiner 330, and the like. The optimal image selecting unit 340 is included.

First, the noise removing apparatus 300 according to an embodiment of the present invention removes or attenuates noise included in a video signal in a block unit, but this is not limited to a block unit and removes noise included in a video signal in a pixel unit. It is also possible.

The current frame / field (hereinafter referred to as 'current frame') (F _n ) and the previous frame / field (hereinafter referred to as 'previous frame') (F _{n -1} ) are the first noise removing unit 310. And a second noise removing unit 320.

The first noise removing unit 310 calculates motion estimation information for each block of the current frame F _n , that is, compares the motion vector v with the previous frame F _{n -1} , and then calculates the motion vector v. The motion estimation information (hereinafter, referred to as a "first candidate image") from which noise is removed is output, and a minimum value of motion prediction error values of each block is output. Here, the motion estimation information is suitable for the video application, and the motion detection information to be described later is suitable for the application of the still image or the background.

In detail, the first noise removing unit 310 divides the current frame F _n into blocks having a predetermined size, and among the divided blocks, the current block and the previous frame F _n-1 which estimate motion information. A plurality of first motion prediction error values are calculated by comparing the search ranges set in the above, and this method is called a block-matching motion estimation technique (BMA).

In this case, the motion prediction error value may be estimated by bi-directional BMA, unidirectional BMA, or all known techniques for estimating a motion vector.

In addition, the motion prediction error value may be calculated by various known methods such as Sum of Absolute Difference (SAD), Mean Absolute Difference (MAD), Mean Square Error (MSE), and the like. SAD value.

The first noise removing unit 310 estimates motion estimation information of the current block from a position having a first minimum SAD value among the calculated first SAD values, and removes noise of the estimated motion estimation information. The image and the first minimum SAD value are output. Here, the removal of noise is performed by various known techniques such as filtering processing, and is not limited to one specific method.

The second noise removing unit 320 compares the previous block corresponding to the same position as the current block among the previous frames F _{n -1 to} determine the motion detection information and the second minimum SAD value. Tell us if it is calculated). The second noise removing unit 320 removes the noise of the calculated motion detection information to obtain a second candidate image, and then outputs a second candidate image and a second minimum SAD value. The first and second noise canceller 320 sequentially performs this process on all divided current blocks and neighboring blocks, and sequentially outputs first and second minimum SAD values of the respective blocks.

The comparison determiner 330 may include the first minimum SAD values of all the divided blocks output from the first noise canceller 310 and the second corresponding to all the divided blocks output from the second noise canceller 320. A reference flag signal for selecting an optimal image is determined by comparing the minimum SAD values.

Referring to FIG. 4, the comparison determiner 330 includes a comparer 332, a memory 334, a first counter 336, a second counter 338, and a determiner 339.

The comparator 332 compares the first and second minimum SAD values calculated from the current block at the same position and outputs a smaller third SAD value. At this time, the comparator 332 compares the first and second minimum SAD values corresponding to all current blocks, respectively, and sequentially outputs the third SAD values of all current blocks.

The memory unit 334 temporarily stores smaller third SAD values sequentially output from the comparator 332 in units of lines. At this time, the number of lines to be stored (n) and the number of third SAD values to be included in the horizontal direction of one line are preferably set in the design of the present invention. In addition, the memory unit 334 may be provided in the same number as the number of lines, or the memory unit 334 may be divided and used by the number n of lines as shown in FIG. 5. The reason for storing the third SAD value in the unit of line as described above is to prevent the output of the unnatural image as shown in FIG. 2 by selecting the optimum image in consideration of the SAD value of not only the current pixel or the current block.

FIG. 5 is a diagram for describing a third SAD value of a line unit stored in the memory of FIG. 4.

4 and 5, a first minimum SAD value S1 is stored in a first cell of a first line, which is a first minimum SAD value and a second minimum value calculated based on the same current block. This means that the first minimum SAD value is smaller among the SAD values. Similarly, the second minimum SAD value S2 is stored in the fourth cell of the third line, and the first minimum SAD value is stored in the last cell. Further, the first minimum SAD value is stored in all cells of the n-th line n-where n is an integer.

The memory unit 334 provides the first minimum SAD value and the second minimum SAD value stored in each line to the first counter 336 and the second counter 338, respectively.

The first counter 336 counts the number of the plurality of first minimum SAD values provided from the memory unit 334, and counts only the first minimum SAD values located in the window w having a predetermined size. Here, the size of the window w is preferably included in (the number n of lines x the number m of SAD values located in the horizontal direction of the line).

The second counter 338 counts the number of second minimum SAD values provided from the memory unit 334, and counts only the second minimum SAD values located in the window w having a predetermined size.

The determination unit 339 determines the reference flag signal based on the counted result and a predetermined threshold value. In more detail, the determining unit 339 may determine whether any one of the number N_1 of counted first minimum SAD values and the number N_2 of counted second minimum SAD values is greater than or equal to a threshold value th, SAD values having a smaller number than the threshold are ignored, and a flag signal corresponding to any one having a larger number than the threshold is determined as the reference flag signal. The determiner 339 outputs the determined reference flag signal to the optimum image selector 340.

That is, the determination unit 339 determines the number of SAD values that are ignored even if the SAD value (for example, the first minimum SAD value) to be ignored is smaller than other SAD values (for example, the first minimum SAD value) at the same position. Is ignored, so ignore it. In other words, the determining unit 339 replaces a small number of SADs with a plurality of SADs if most of the regions within a given window are defined as a specific SAD (that is, one of blocks related to motion estimation information or SADs related to motion detection information). To match the characteristics of the window.

For example, as shown in FIG. 5, if the number N_1 of the first minimum SAD values located in the predetermined window w is 14, the number N_2 of the second minimum SAD values is 2, and the threshold value is 5, the determination is made. Since the number N_1 of the first minimum SAD values is greater than the threshold value, the unit 339 determines the first minimum SAD value, that is, the flag signal '0' corresponding to the motion estimation information, as the reference flag signal. The determination unit 339 outputs the reference flag signal '0' to the optimum image selecting unit 340. That is, the determination unit 339 ignores two second minimum SAD values smaller than the first minimum SAD value, and gives high reliability to the first minimum SAD value having a high ratio.

On the other hand, if the number N_2 of counted second minimum SAD values is greater than the threshold value, the determination unit 339 selects the reference flag signal '1' corresponding to the second minimum SAD value, that is, the motion detection information. Output to the unit 340.

The optimum image selector 340 selects and outputs one of the first and second candidate images as an optimal image based on the reference flag signal output from the determiner 339.

In detail, the first candidate image generated by the first noise remover 310 and the second candidate image generated by the second noise remover 320 are buffers (not shown) provided in the optimum image selector 340. Temporarily stored in

When the flag signal '0' is input from the determiner 339, the optimum image selector 340 selects an image corresponding to '0', that is, a first candidate image as the optimum image. This means that the flag signal '0' means the first minimum SAD value, and since the first minimum SAD value and the first candidate image are output from the first noise canceller 310, the optimal image selector 340 is set to '0'. The first candidate image corresponding to 'is selected as the optimal image.

In addition, when the flag signal '1' is input from the determiner 339, the optimum image selector 340 selects an image corresponding to '1', that is, a second candidate image as the optimal image. This means that the flag signal '1' means the second minimum SAD value, and since the second minimum SAD value and the second candidate image are output from the second noise canceller 320, the optimal image selector 340 may be '1'. The second candidate image corresponding to 'is selected as the optimal image.

On the other hand, if both the number N_1 of the counted first minimum SAD values and the number N_2 of the counted second minimum SAD values are smaller than the threshold value, the determination unit 339 checks the memory unit 334 to determine the current state. The flag signal of the SAD value corresponding to the block is output to the optimal image selector 340 as a reference flag signal.

FIG. 6 is a flowchart schematically illustrating an image processing method for noise reduction implemented by FIG. 3.

3 to 6, the first noise removing unit 310 compares a current block of a current frame with a predetermined search area of a previous frame, calculates a first minimum SAD value and motion estimation information of the current block, and moves the motion. The first candidate image is generated by removing noise of the estimated information (S605).

The second noise removing unit 320 compares a previous block corresponding to the same position as the current block among the previous frames F _{n -1} , calculates motion detection information and a second minimum SAD value, and removes noise of the motion detection information. In operation S610, the second candidate image is generated by removing the second candidate image. Here, the first and second minimum SADs are provided to the comparator 332, and the first and second candidate images are provided to the optimum image selector 340.

When operation S610 is performed, the comparator 332 compares the first and the minimum 2SAD values calculated from the current block at the same position and outputs a smaller SAD value (3SAD) (S615).

After operation S615, the memory unit 334 temporarily stores smaller SAD values sequentially output for each current block from the comparator 332 in units of lines, and stores the first minimum SAD value and the second minimum SAD stored in each line. Values are provided to the first counter 336 and the second counter 338, respectively (S620). In this case, it is preferable that the number (n) of the lines stored in step S620 and the number of the third SAD values to be included in the horizontal direction of one line are preset.

The first counter 336 counts the number N_1 of the first minimum SAD values provided from the step S620, but counts only the first minimum SAD value located in the window w of the predetermined size, and the second counter 338. ) Counts the number N_2 of second minimum SAD values provided from step S620, and counts only the second minimum SAD values located in the window w of a predetermined size (S625).

When the step S625 is performed, the determination unit 339 determines that any one of the counted first minimum SAD and the second minimum SAD is equal to or greater than a predetermined threshold value (S630). The image is determined by the optimal image selector 340 (S635).

The optimal image selector 340 selectively outputs one of the first and second candidate images as an optimal image based on the provided reference flag signal (S640). For example, when the reference flag signal '0' is provided, step S640 selects the first candidate image corresponding to '0' as an optimal image. When the reference flag signal '1' is provided, step S640 is the second candidate. Select the image as the best image.

On the other hand, if it is determined in step S630 that both the number of the first minimum SAD and the number of the second minimum SAD are smaller than the predetermined threshold value, the determination unit 339 corresponds to the current block among the plurality of third SAD values stored in step S620. The SAD value is checked (S645), and the reference flag signal corresponding to the confirmed SAD value is output to the optimum image selector 340 (S650).

According to an image processing apparatus capable of noise reduction according to the present invention and an image processing method for noise reduction thereof, the SAD values of neighboring blocks or pixels as well as current blocks or pixels are filtered in processing for removing noise to select an optimal image. By selecting the optimal image, a part of the screen can be extremely unobtrusive or unnatural.

In particular, by selecting the optimal image in consideration of the surrounding SAD values, it is possible to prevent the situation in which the image information by the motion detection is selected in the region where the image information by the motion estimation is to be selected as the optimal image or vice versa and prevent the natural screen Can be implemented.

Although the present invention has been described in detail through the representative embodiments, those skilled in the art to which the present invention pertains can make various modifications without departing from the scope of the present invention with respect to the embodiments described above. I will understand. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined by the claims below and equivalents thereof.

Claims (5)

Comparing the current block of the current frame with a predetermined search area of the previous frame to obtain a first minimum motion prediction error value and motion estimation information of the current block, and removing noise of the motion estimation information to output a first candidate image 1 noise removing unit; Comparing a previous block corresponding to the same position as the current block among the previous frames, a second minimum motion prediction error value and motion detection information of the current block are obtained, and a second candidate image is removed by removing noise of the motion detection information. An output second noise removing unit; A comparison determination unit determining a reference flag signal for selecting an optimal image by comparing the output first and second minimum motion prediction error values with a predetermined threshold value; And And an optimal image selector configured to selectively output one of the first and second candidate images as the optimal image based on the determined reference flag signal. The method of claim 1, The comparison determination unit, A comparator for outputting a smaller value among the first and second minimum motion prediction error values; A memory unit which stores the output small value in line units; A counter for counting each of the first and second minimum motion prediction error values included in any window of a predetermined number of lines stored in the memory unit; And And a determination unit to determine the reference flag signal based on the counted result and the predetermined threshold value. The method of claim 2, The determining unit outputs a reference flag signal corresponding to any one of the counted first and second minimum motion prediction error values to the optimal image selecting unit when the number of the counted first and second minimum motion prediction error values is greater than or equal to the threshold value. And the optimum image selecting unit selects a candidate image corresponding to the output reference flag signal among the first and second candidate images as the optimal image. (a) obtaining a first minimum motion prediction error value and motion estimation information of the current block by comparing the current block of the current frame with a predetermined search area of the previous frame, and removing the noise of the motion estimation information to obtain a first candidate image. Outputting; (b) comparing a previous block corresponding to the same position as the current block in the previous frame to obtain a second minimum motion prediction error value and motion detection information of the current block, and removing noise of the motion detection information to remove the second block; Outputting a candidate image; (c) determining a reference flag signal for selecting an optimal image by comparing a predetermined threshold value of the number of output first and second minimum motion prediction error values; And and (d) selecting and outputting one of the first and second candidate images as the optimal image based on the determined reference flag signal. The method of claim 4, wherein In step (c), (c1) outputting a smaller value of the first and second minimum motion prediction error values; (c2) storing the output small value in units of lines; (c3) counting each of the first and second minimum motion prediction error values included in any window of the stored predetermined number of lines; And and (c4) outputting a reference flag signal corresponding to any one of the counted first and second minimum motion prediction error values when the number of the counted values is equal to or greater than the predetermined threshold value. In the step (d), the candidate image corresponding to the flag signal output in the step (c4) among the first and second candidate images is selected as the optimal image.

Images