KR20040105274A - Method for blocking artifact reduction included ended image based on bdct - Google Patents

Abstract

PURPOSE: A method for removing a blocking artifact included in an image encoded by BDCT(Block Discrete Cosine Transform) is provided to effectively reproduce a received image by removing the blocking artifact without deleting the edge of the received image and to reduce the quantity of calculations for removing the blocking artifact to process the received image in real time. CONSTITUTION: An image encoded by BDCT is received. The frequency characteristic of the received image is measured. The received image is moved by at least one pixel in column and row directions, and the frequency characteristic of the moved image is measured. When the difference between the frequency characteristic of the received image and the frequency characteristic of the moved image exceeds a predetermined range, a high-frequency component of the image is removed.

Description

METHODO FOR BLOCKING ARTIFACT REDUCTION INCLUDED ENDED IMAGE BASED ON BDCT}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video encoding system, and more particularly, to a method of restoring an edge component while removing a block phenomenon present in a video signal.

The need for compression on digital images is growing regardless of whether they are still images or moving images (i.e. video images). Most still image and video encoding standards currently used internationally use partition unit processing and variable length coding to obtain high data compression efficiency. These systems are very sensitive to errors that occur when retrieving or transmitting compressed data. This is because an error of one bit occurring in the encoded bitstream may affect the entire block without being limited to one pixel in the reconstructed image data.

Block discrete cosine transform (BDCT) is widely used in still and video coding standards such as JPEG, H.261, and MPEG. In this technique, an image is generally divided into 8 × 8 blocks without overlapping, and the divided blocks are independently encoded without considering correlation between blocks. In other words, the video compression transmission techniques such as JPEG and MPEG divide the video into blocks and perform compression. Thus, a small amount of compressed video appears as if mosaic processing is performed on a TV. It is called (Blocking Artifact). Hereinafter, the encoding and decoding process by the BDCT will be described.

1 illustrates a general encoding and decoding process by BDCT. Hereinafter, a general encoding and decoding process by the BDCT will be described with reference to FIG. 1. FIG. 1 is McGraw-Hill, Inc. in 1995 by Daniel Minoli. "Video Dialtone Technology", published in. In the encoding process, the input image 100 is first decomposed into blocks of 8 × 8 pixels, and the DCT operation 102 is performed for each block. When the DCT operation is performed, the entire area is decomposed into a direct current (DC) component and an alternating current (AC) component. The decomposition process is called an orthogonal transformation.

Various orthogonal transformations have been proposed since the 1960s, among which DCT, proposed by Professor Rao of the University of Texas, is considered the most effective method for image compression. The DCT proposed by Professor Lao's team has been adopted in international standards such as MPEG, JPEG, and H.261. An advantage of the orthogonal transformation is that a pixel value that is irregularly spread on the screen before the conversion tends to concentrate toward the low frequency after the conversion. Therefore, it is possible to perform information compression with little information loss by manipulating the high frequency terms. Thus discarding high frequency terms is quantization. The quantization is to remove the pixel value or the component value of each frequency orthogonally transformed by the quantizer 104 by the quantization coefficient QSF stored in the Huffman table 120 to round off the remainder. Therefore, the part lost by the above process cannot be completely restored to the original state even if the quantization coefficient 120 is multiplied when the image is restored. The larger the value of the quantization coefficient, the higher the compression ratio, but because the quantization error is larger, the decoded image 110 is blurred. The smaller the value of the quantization coefficient is, the smaller the portion that is cut off becomes. The closer to the original image 100, but the compression rate is lowered. The DC coefficients among the DCT coefficients of the quantized image are encoded by an entropy encoder 106 using the DC coefficients of the preceding blocks as prediction values. The remaining AC components are arranged in a line by zigzag scan for each block and then encoded. Through this encoding process, compressed image data 124 is obtained.

The decoding process is the reverse of the encoding process. First, the compressed image data 124 is inversely quantized by using the Huffman table 122 and the quantization table 120 by an entropy decoder 116 and a quantizer 114. The decoded image 110 is configured by using the inverse DCT transform (IDCT) 112 along with the decoded DC coefficients. Since each block of 8x8 pixels is independently encoded, a block phenomenon occurs because a DC value drops in each block or a seam of blocks becomes discontinuous.

Methods for removing the block phenomenon include a technique using a linear low band filter, a technique using a non-linear spatial adaptive filter, and finally an iterative approach based on Projection onto Convex Sets (POCS). The technique using the linear low pass filter is most commonly used due to the simplicity of implementation. In the technique using the linear low band filter, as described above, since the images compressed by the encoding process have low frequency components, a high frequency component generated by a block phenomenon is removed using a low frequency filter. However, the linear low-band filter has a disadvantage in that the image quality is reduced by blurring an edge component, which is an image signal having a high frequency component. To solve this problem, a more complex nonlinear spatial adaptive filter has been proposed. This technique does not simply remove high frequency components, but removes the block phenomenon by using a filter most suitable for each block in consideration of the pattern and edge pattern in which the block phenomenon occurs. However, there is a limit to preserving the edge of the image even in the scheme using the nonlinear spatial adaptive filter.

The POCS-based repetition technique is more efficient in removing block phenomena while preserving edges of an image, compared to a method of removing block phenomena using a filter. However, this technique performs several processes that repeat the DCT and IDCT processes. Due to the above disadvantages, a large amount of calculations cannot be used in a video communication system that transmits data in real time. Therefore, there is a need for a method capable of effectively removing block phenomena without adversely affecting the sharpness of image data.

Accordingly, an object of the present invention is to propose a method of removing the block phenomenon from an image signal including the block phenomenon.

Another object of the present invention is to propose a method for efficiently reconstructing a received image by removing the block phenomenon without deleting an edge portion which is a part of the received image.

It is still another object of the present invention to propose a method of processing a received image in real time by reducing a computation amount for removing a block phenomenon.

In order to achieve the above object of the present invention, an image generated by a block discrete cosine transform (BDCT) method of dividing a single image into a plurality of blocks and compressing each of the plurality of blocks is received. A method for removing an included block phenomenon, the method comprising: measuring a frequency characteristic of a received image, moving the received image in row and column directions by at least one pixel, and Measuring the frequency characteristics, and if the difference on the characteristics of the measured frequencies exceeds the set range, the process of removing the high frequency components.

In order to achieve the above object of the present invention, by comparing the DCT of the received image with the position where 0 appears in the DCT of the image shifted in at least one pixel row and column direction, whether the block phenomenon of the received image occurred It is a process of judgment.

1 is a diagram illustrating a process of encoding / decoding an image to be transmitted by a block discrete cosine transform scheme.

2 is a diagram illustrating a process of removing a block phenomenon from an image including the block phenomenon.

3 is a diagram illustrating a process of measuring frequency characteristics by discrete cosine transform of a received image and an image shifted by a predetermined pixel according to the present invention;

4 is a diagram illustrating a process of scanning in zigzag.

5 is a diagram illustrating a process of obtaining a discrete cosine transform of an image shifted by a predetermined pixel by using a received image according to the present invention.

6 is a diagram illustrating a process of removing a block phenomenon from an image including the block phenomenon according to the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. In describing the present invention, when it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

2 illustrates a process of receiving and processing a video signal having a block phenomenon according to the present invention. Hereinafter, a process of processing the received video signal to which the present invention is applied will be described in detail with reference to FIG. 2.

The received video signal Has a block phenomenon. As described above, the block phenomenon refers to a seam of discontinuous blocks when the image signal is processed into blocks having a constant size. The higher the compression rate, the larger the block phenomenon, and the lower the compression rate, the smaller the block phenomenon. As described above, the block phenomenon and the compression rate have a correlation with each other. The video signal with the block phenomenon has two limitations. One of the two constraints is smoothness constraint sets ( ) And the other one is quantization constraint sets ( ) Is projected. remind The projection of is shown in Figure 2 above Project against, and The projection of is shown in Figure 2 above Project against. That is, as shown in FIG. 2, the block phenomenon of the video signal is Remind Project and then perform the projection process. Remind Project on. remind Is a process of removing high frequency components from a video signal having a block phenomenon. By performing the process of removing the high frequency component, the POCS not only removes the block phenomenon but also simultaneously removes edge components included in the video signal. In order to overcome the unreasonable point of removing the edge components at the same time as described above Perform the process. remind The process not only partially restores the removed edge components but also partially removes the removed block phenomenon components. By repeating the above process, the signal from which the block phenomenon is removed ( ) The above Wow Learn about.

remind Is carried out in IDCT and said Is performed in the DCT. Therefore, in order to use the iterative technique based on the POCS theory, the IDCT process and the DCT process should be repeated. In addition, as described above Limits the frequency band of the image in the vertical and horizontal directions to remove high frequency noise in the vertical and horizontal directions. That is Is a condition for removing the block phenomenon, and the edge of the original signal is also blurred. In order to overcome this blurring of the edge of the original signal, Determine.

3 shows an 8 × 8 block, which is a unit in which an image is compressed, and a block delayed by a predetermined interval. Hereinafter, the present invention is applied using FIG. 3. Learn about the process of obtaining. FIG. 3 determines whether a block phenomenon occurs in a received image by measuring a correlation between the received image and the received image and the delayed image.

3 is an image compressed in 8 × 8 block units And above Is moved one pixel each in the row and column direction. It is shown. Although FIG. 3 illustrates a case in which the pixels are moved in the row and column directions by one pixel, they may be moved by at least one pixel unit according to a user's selection. remind Above Process for generating the above Above This is the process to create. As shown in FIG. 3, there is a constant correlation between the 8 × 8 block and the 8 × 8 block delayed by one pixel. Thus, the DCT coefficient and The DCT coefficients of V are similar because they have a difference of one pixel. If there is a difference, it may be due to the influence of the high frequency component caused by the block phenomenon. In FIG. 3, the frequency characteristics of the original block and the one-pixel shifted block are analyzed. Wow Suggest.

An input image of mN × nN is defined as f, and a pixel for the input image is defined as f (a, b). (A, b) represents the coordinates of the pixel. That is, (3, 2) means a pixel located in the third to the left and the second to the top of the input image composed of a plurality of pixels. Also, the And above Is N × N vectors representing N × N elements of each block, and is represented by Equation (1).

As shown in FIG. Is the original block Moved by one pixel at, It includes the boundaries of the right and bottom of the. That is It includes the block phenomenon that occurs at the block boundary of. remind And above If the difference between the values is greater than or equal to a certain size, it is recognized that a block phenomenon occurs at the block boundary. But remind And above If the difference in the values for is below a certain magnitude, i.e. And above If there is no difference in the value of, it is recognized that no block phenomenon occurs at the block boundary. Remind to simplify the expression Wow Is expressed as b and s, and the DCT result of b (x, y) and s (x, y) is B (u, v) and S (u, v). .

In general, an image to be transmitted is encoded by dividing it into 8 × 8 blocks and compressing it again. That is, it can be expressed as <Equation 3>.

If the position of non-zero coefficient of B is farther than the position of non-zero coefficient of S when scanned in zigzag order in the decoded image, these coefficients can be recognized as a sudden change in brightness between blocks. . That is, it can be recognized that a block phenomenon occurs at the block boundary. Figure 4 shows the scanning process in the order of zigzag according to the present invention. In order to reduce the block phenomenon by the above concept Wow Suggest. (m-1) of (n-1) Is defined as in Equation 4.

NZ (·) represents the last position where the non-zero DCT coefficients appear in zigzag order, and T ( ) Denotes the last position of the DCT coefficient of S, which is left to preserve the naturalness of the image. remind Above Is an estimate of the variance of and is calculated as the sum of the absolute values of the first five AC coefficients of S. small Block with Can be thought of as an area with no change. remind Since T has a small value, the T ( ) Also has a small value, indicating that many high frequency components have been removed. On the other hand Block with May be determined as an area including an edge, and T ( ) Preserves the edges.

The above from Learn about the process of finding. Predetermined NZ ( ) + T ( Farther away We can get all the nonzero DCT coefficients of to 0 to get the projected element. That is, high frequency components due to the block phenomenon are removed.

Based on the above description, the present invention proposes a method for processing an input image in real time by eliminating DCT and IDCT processes and simplifying a complicated calculation process. Hereinafter, a process of obtaining a DCT for an image moved by a set number of pixels using the DCT of the received image will be described. By performing the above process, it is possible to reduce the amount of computation for calculating the DCT for the image moved by the set number of pixels. As shown in FIG. 5, one block including a part of each of the four blocks may be obtained as shown in Equation 5. The four blocks represent a received image, and one block including a portion of each of the four blocks represents an image moved by a set number of pixels.

here, , , And above Wow Is a matrix of size hxh and wxw, respectively.

Since DCT performs linear orthogonal transformations, the distribution law holds for matrix multiplication, Of DCT Can be obtained as shown in Equation 6.

remind Wow Is Wow Is DCT. Using the above method, it is possible to convert from B (u, v) to S (u, v), or from S (u, v) to B (u, v). That is, the DCT of the other image can be calculated by calculating the DCT of one of the two images without calculating the DCT of the received image and the DCT of the image moved in the row and column directions by the set number of pixels. have.

6 illustrates a process of receiving a video signal including a block phenomenon according to the present invention and removing the block phenomenon from the received video signal. As shown in FIG. 6, two constraints, smoothness constraint sets ( ) And the quantization constraint sets ( It can be seen that the process of projecting) is performed in the DCT region.

As described above, the present invention efficiently restores an image by removing the block phenomenon generated by the block discrete cosine transform method in real time. In addition, by removing only the high-frequency components due to the block phenomenon and not the edge components that are part of the image, the same image as the transmitted image is restored.

Claims (7)

Receiving an image generated by a block discrete cosine transform (BDCT) method that divides one image into a plurality of blocks and compresses each of the plurality of blocks, and removes a block phenomenon included in the received image. In the method, Measuring a frequency characteristic of the received image; Moving the received image by at least one pixel in a row and column direction, and measuring a frequency characteristic of the moved image; And removing a high frequency component if the difference of the characteristics of the measured frequencies exceeds a set range. The method of claim 1, wherein the frequency characteristic of the received image and the frequency characteristic of the image shifted in a row and column direction by at least one or more pixels are measured by a discrete cosine transform (DCT) method. The method of claim 2, wherein the discrete cosine transform method is measured by Equation 7 below. B (u, v): DCT for received image S (u, v): DCT for image shifted in at least one pixel row and column direction 4. The method of claim 3, wherein a block phenomenon of the received image is determined by comparing a DCT of the received image with a position where 0 appears in the DCT of the image shifted in at least one pixel row and column direction. The method characterized in that. The method as claimed in claim 4, wherein the position where the zero is represented is determined in proportion to the difference, and the block phenomenon of the received image is generated. 4. The method as claimed in claim 3, wherein the step of reducing the block phenomenon by removing the high frequency component is projected onto a projector generated by Equation (8). The method as claimed in claim 2, wherein the DCT for the image shifted by at least one pixel in the row and column directions is obtained by the following Equation (9). : DCT DCT