KR100997218B1 - Image dequantization method - Google Patents

Abstract

The image dequantization method includes decoding quantization information and additional quantization information of a magnetic region included in one slice unit header from a bit string, and including the decoded quantization information and additional quantization information in the first region. Determining a first quantization coefficient of the extracted macroblocks, obtaining a second quantization coefficient of the macroblocks included in the second region through the decoded quantization information of the magnetic domain, and using the first quantization coefficient Performing inverse quantization of the macroblocks included in the first region, and performing inverse quantization of the macroblocks included in the second region by using the second quantization coefficient.

Description

Image dequantization method

The present invention encodes and decodes a luminance component and a color component using different quantization methods in consideration of visual characteristics of a person who is sensitive to color change between regions when encoding and decoding using different quantization coefficients for each region in one image. It is about a method.

In general, when encoding a single image, encoding is performed by dividing the luminance component and the color component separately. This is because the amount of information and statistical characteristics of each component are different. MPEG-4 part 2 standard, the international standard of moving picture (ISO / IEC 14496-2, Information Technology - generic Coding of Audio - Visual Objects , Part 2: Visual ( Final Draft of International Standard ) , Dec. 1998 and H.263 version 2 standard (ITU-T Recommendation H.263, Video) Coding for Low Bit Rate Communication , Jan. In 1998, color images are encoded by dividing them into luminance components (Y) and color components (chroma; U, V). Fig. 3 shows the configuration of a macroblock used as a basic unit in video compression in MPEG-4 and H.263 standards. It is composed of four 8x8 Y blocks and one 8x8 U and V block, respectively, with luminance component Y and color. It can be seen that components U and V are separately encoded.

The MPEG-4 and H.263 standards, on the other hand, allow encoding and decoding using different quantization coefficients for different regions when encoding a single image. In this case, there is a case where a slice structure is used and difference quantization is used to control the bit rate for each macro block.

First, in the case of using a slice structure in H.263, when encoding a single image, a GOB layer is formed in units of macroblock columns, and the quantization coefficients used in the GOB are encoded in each GOB header. In Annex K, one image may be encoded by dividing a region into a group unit of consecutive macroblocks or a macroblock group of an arbitrary rectangular shape unit using a slice structured mode.

4A illustrates an example in which slices are configured in units of three GOBs, and FIG. 4B illustrates an example in which slices are configured in units of squares. Each slice can be independently decoded with a slice header. The slice header starts with the start code and has a position in the image of the starting macroblock, the size of the slice, and a quantization coefficient used to encode the macroblock in the slice. . 5A and 5B show an example of a slice header.

As such, when the slice structure is used, different quantization coefficients may be used for each slice. In particular, when the rectangular slice structure is used as shown in FIG. 4B, the background region and the ROI are distinguished, and a small quantization coefficient value is used for the ROI. The image quality is increased and the background region is compressed using a large quantization coefficient value to increase the overall subjective image quality under a limited bit rate.

In MPEG-4, bit rate control can be performed in units of macroblocks. To this end, bit rates are adjusted for each macroblock by changing a quantization coefficient by adding or subtracting a specific value to the quantization coefficient for each macroblock. The specific value used at this time is called a difference quantization value and is encoded for each macroblock.

As such, when different quantization coefficients are used for each region in a single image, luminance and color values are different for each region. If different colors appear in each region, the subjective image quality of the image is impaired.

In the present invention, when the quantization is performed using different quantization coefficients for each region in a single image, the amount of change in the quantization coefficient is changed smaller in the color component than in the luminance component to enhance the subjective image quality of the person.

An object of the present invention is to increase the subjective image quality of a person by changing the amount of change in the quantization coefficient for each region smaller than in the luminance component when encoding the image by region.

Another object of the present invention is to provide a recording medium recorded with a program code executable by a computer in an encoding and decoding method using the color quantization method.

In order to achieve the above object, the image encoding method according to the present invention comprises the steps of (a) setting the initial value of the quantization coefficients for dividing the input image input by the unit of the screen into a region of a constant size and quantizing and encoding each region; Each region set in step a) is divided into independent slice structures, and encoding is performed in each slice header including location information of the region and quantization coefficients to be used in the corresponding region or the adjacent region. In order to encode each area by a unit of block of a certain size, the quantization coefficients of each block are processed differently for the luminance component and the chromaticity component, and the quantization coefficients of each block are adjusted according to the position or bit rate of the region so that the color variation between the regions is small. In step (c) of determining and predicting / transforming encoding the image data for each block, according to the quantization coefficient determined in step (c) Quantization to be used for encoding a subsequent block in order to control the bit rate by using the bit amount of the bit stream generated in (d) and (e) encoding the entropy of the quantized image data in units of blocks. Step (f) is calculated and passed to step (c).

In order to achieve the above object, the image encoding apparatus according to the present invention divides an input image input in units of a screen into regions having a predetermined size, and includes an area modeling unit and an area modeling unit for setting initial values of quantization coefficients for quantizing and encoding each region. A slice header encoder which separates each region set by the sub-part into an independent slice structure and encodes each slice header including location information of the region and quantization coefficients to be used in the corresponding region or the adjacent region, and blocks each region by a predetermined size. The quantization coefficient modeling unit for each block determines the quantization coefficient of each block according to the position or the bit rate of the region so that the quantization coefficient of each block is processed differently for the luminance component and the chromaticity component for encoding with Region after prediction / conversion encoding In order to control the bit rate by using the bit rate of the bit stream generated by the entropy encoder and the entropy encoder, which is quantized according to the quantization coefficients determined by the quantization coefficient modeling unit, the quantization coefficient to be used for encoding the block is calculated and then And a bit rate control unit which is passed to the star quantization coefficient modeling unit.

In order to achieve the above object, in the image decoding method according to the present invention, in the step (g) of decoding the slice header into the received bit stream, the entropy decoding of the received bit stream for each slice (h), (h) step From the decoded information, the quantization coefficients of the respective regions are determined using the quantization coefficients of the regions included in the slice header so that the color change between the regions is small. (J) reconstructing the image through quantization / inverse transform / prediction compensation; And (k) configuring the image of each region reconstructed in step (j) into one screen by referring to the position information of each region included in the slice header.

In order to achieve the above object, an image decoding apparatus according to the present invention includes a slice header decoder for decoding a slice header into a received bit string, an entropy decoder for entropy decoding a bit string received for each slice, and a decoder decoded in an entropy decoder. Using the quantization coefficient modeling unit for each region and the quantization coefficient modeling unit for determining the quantization coefficient of each region so that the color change between regions is small by using the quantization coefficients of each region included in the slice header among the information. An inverse quantizer for inverse quantization, an inverse quantizer for inversely quantizing the inverse quantized signal, and an image restoring unit for restoring an image by compensating a predicted signal; Image composed of one screen by referring to location information of area It includes a component.

According to the method and apparatus for image encoding and decoding through color quantization in consideration of the visual characteristics of a human, according to the present invention, the amount of change in the quantization coefficient used for each region when encoding or decoding a single image by region is determined by the luminance component. The smaller the chromatic component, the smaller the color difference between regions, thereby increasing the subjective quality of the reconstructed image. This may be applied to dividing each region into independent slice structures, encoding or decoding each region using different quantization coefficients, or changing the quantization coefficients to match the bit rate within one region.

Hereinafter, a method and an apparatus for encoding and decoding an image through color quantization in consideration of a visual characteristic of a person according to the present invention will be described in detail with reference to the accompanying drawings.

1A is a block diagram schematically illustrating an embodiment of an apparatus for encoding a region of interest according to a visual characteristic of a person, according to an embodiment of the present invention. The region modeling unit 100, the slice header encoding unit 101, and the quantization coefficient modeling for each region are shown. A unit 102, a spatial / temporal predictive encoding unit 103, a transform encoding unit 104, a quantization unit 105, an entropy encoding unit 106, a bit rate control unit 107, and a bit string generator 108 It is configured by.

FIG. 1B is a block diagram schematically illustrating an embodiment of an apparatus for decoding a region of interest according to a visual characteristic of a person according to an exemplary embodiment of the present invention, wherein a slice header decoder 114, an entropy decoder 109, and quantization coefficient modeling for each region are shown. The unit 110 includes an inverse quantization unit 111, an image restoration unit 112, and an image configuration unit 113.

2A is a flowchart illustrating an embodiment of a region of interest image encoding method performed by the apparatus of FIG. 1A, and FIG. 2B is a flowchart illustrating an example of a region of interest image decoding method performed by the apparatus of FIG. 1B.

A preferred embodiment of the encoding method of the present invention will be described with reference to FIGS. 1A and 2A. The region modeling unit 100 sets various types of regions such as a region of interest, a background region, and a variation region in order to divide and encode an input image input in a screen unit into a slice structure, and then initial values of quantization coefficients to be used for each region. In operation 200, the slice header encoder 101 encodes a value required for each region, such as the position of each region and the value of the quantization coefficient, received from the region modeling unit 100 in the slice header (S201). step).

The region quantization coefficient converter 102 sets the value of each region quantization coefficient by referring to the initial value of the region quantization coefficient set by the region modeling unit 100 (step 202).

The predictive encoding unit 103 predictively encodes the input video according to the information of each region received from the region modeling unit 100 for each slice and predicts and encodes the image in space and time, and the transform encoding unit 104 predictively encodes the input image. Is converted through a conversion such as DCT (step 203). On the other hand, in order to simplify the encoding calculation, the predictive encoding unit 103 and the transform encoding unit 104 perform encoding by dividing the area into block units having a predetermined size.

The quantization unit 105 receives the prediction encoded and transform-coded image data for each block and quantizes the image data for each block according to the quantization coefficients determined by the region-based quantization coefficient converter 102 (step 204).

The entropy encoder 106 encodes the entropy of the image data quantized by the quantizer 105 and outputs the encoded entropy to the bit string generator 108 (step 205).

The bit rate controller 107 refers to the bit amount of the bit string output from the entropy encoder 106, adjusts the value of the quantization coefficient so as to adjust the bit amount of a block to be encoded later, and converts this information into a quantization coefficient converter for each region ( 102) (step 206).

Meanwhile, the bit string generator 108 adds the slice header data received from the slice header encoder 101 to the entropy encoded image data received by the entropy encoder 106 to generate a bit string.

A preferred embodiment of the ROI decoding method of the present invention will be described with reference to FIGS. 1B and 2B. The decoding apparatus of the present invention receives a bit string and stores it in a buffer (step 207). The slice header decoder 108 decodes values required for each region, such as the position of each region included in the slice and the value of a quantization coefficient. (Step 208). The position value of each region is used to construct a screen and reconstruct data in the image forming unit 113, and the value of the quantization coefficient of each region is used to set the quantization coefficient for each region in the region-based quantization coefficient modeling unit 110. do.

The entropy decoding unit 109 decodes a bit string of data belonging to each region (step 209).

The quantization coefficient modeling unit 110 for each region determines the value of the quantization coefficient for each block of the corresponding region by referring to the shape of the slice decoded by the slice header decoder 108, the position and size of each region, and the quantization coefficient value. do. In this case, the values of the quantization coefficients may be gradually changed over a predetermined period so as not to change rapidly between the respective regions.

The quantization coefficient modeling unit 110 for each region determines the value of the quantization coefficient by reflecting the information when there is a difference quantization value among the information decoded by the entropy decoder 109.

When the quantization coefficient modeling unit 110 sets the value of the quantization coefficient in each region, if the value of the quantization coefficient changes in each region or block, the value of the quantization coefficient is set so that the variation of the color component is smaller than the variation of the luminance component. Reduce the color difference. This is because the subjective picture quality deteriorates when the color changes for each region in an image.

The inverse quantization unit 111 dequantizes the block data received by the entropy decoding unit 109 using the values of the quantization coefficients determined by the region-based quantization coefficient modeling unit 110 (operation 210). In operation 112, an inverse transformation is performed for each block, and the image is restored by compensating the predicted information (step 212). In the image configuring unit 113, the image reconstructed for each region is combined according to a corresponding position to form a screen. (Step 213).

The method of setting the section in which the quantization coefficients are to be changed or the method of changing the quantization coefficients at the time of decoding uses the same method as in encoding.

As described above, when the values of the quantization coefficients are changed between each region or between blocks in each region when encoding or decoding the regions by dividing the regions into one slice in each image, the change in the luminance component and the chromaticity component is changed. In other words, it is possible to effectively eliminate the phenomenon of color difference between regions while increasing coding efficiency.

FIG. 3 shows the configuration of macroblocks used as basic units when compressing images in MPEG-4 and H.263 standards. It is composed of four 8x8 Y blocks and one 8x8 U and V block, respectively. It can be seen that the color components U and V are separately encoded. At this time, since the information amount of the luminance component is larger than the information amount of the color component, the compression ratio greatly depends on the value of the quantization coefficient of the luminance component. Therefore, by changing the value of the quantization coefficient for each region, the compression efficiency can be increased or the compression can be performed according to the desired bit amount. In this case, in the conventional method, the value of the quantization coefficient of the color is also changed. When the color is changed within a certain region of an image, the subjective image quality of the image is damaged by different colors. Therefore, it is possible to improve the subjective picture quality of the image by removing the phenomenon in which the color difference occurs between regions by making the change amount of the quantization coefficient used for color smaller than the change amount of the quantization coefficient used for luminance.

4A and 4B show an example in which a region is set to divide and encode an input image into a slice structure. In the case of Figure 4a shows a slice structure consisting of a group of block (GOB) unit consisting of a row of macroblocks commonly used in H.263, 400, 401, 402 each form a slice. The quantization coefficients to be used for each slice are encoded in the slice header so as to have different quantization coefficients between the slices. In this case, if the quantization coefficients are different between the slices, due to the difference in the quantization coefficients, the luminance difference and the chromaticity difference between the respective slices become clear when the encoded image is restored, thereby creating a horizontal boundary. 5A illustrates this slice header.

In the case of FIG. 4B, the slice structure is configured by setting a rectangular area in one image. In this case, by setting a region of interest 404, which is a part that the user wants to see in more detail, is a rectangular region, the image quality of this region is increased, and the background region 405, which is the other region, is increased in compression to increase the overall subjective image quality. Can be improved. This method can be implemented using small quantization coefficients in the region of interest and large quantization coefficients in the background region. However, in this case, since the boundary may appear prominently due to a sudden change in image quality between the two regions, the variation region 403, which is a section in which the quantization coefficient may change gradually, may be set. As described above, in order to gradually change the quantization coefficients, not only the quantization coefficients used in the slice but also the quantization coefficients of adjacent regions must be encoded together in the slice header. 5B is an example of such a slice header.

In this way, the variation region 403, which is a section in which the quantization coefficients can gradually change between the ROI and the background region, is set, and when the quantization coefficient is changed, the amount of change in the quantization coefficient of the color component is reduced so that no color difference occurs between the regions. There is a need. A method of determining the amount of change in the quantization coefficient of the color component with respect to the amount of change in the quantization coefficient of the luminance component is described below.

6 is a diagram showing how the relationship between the quantization coefficients of the luminance component and the chromaticity component is set when this value is given to the header. As shown in FIG. 6, the luminance component has a one-to-one correspondence with the quantization coefficient of the header. However, as the quantization coefficient increases, the chroma component generally has a smaller size than the luminance component. This is to reduce the sensitivity to chromatic components in the encoding because it affects the human eye more. The relationship between the quantization coefficients shown in FIG. 6 is basically applied to the quantization structure of the present invention.

FIG. 7 illustrates a method for reducing the difference in luminance and color components caused by using different quantization coefficients between the ROI 404 and the background region 405 in the ROI encoding and decoding scheme. In the slice header, the quantization coefficients of the region of interest 404 and the background region 405 are obtained (701), and the quantization coefficients of the luminance and chromatic components of the region of interest 404 and the background region 405 are obtained using these two quantization coefficients. Obtain (702). At this time, the quantization coefficients for the luminance and chromaticity components of each region have a relationship as shown in FIG. 6. The difference d ₁ between the quantization coefficients between the background region 405 and the region of interest 404 obtained in operation 701 is obtained, and the degree of visual difference between the two regions is calculated (703).

If the quantization coefficient QPb of the background region 405 obtained in step 701 is smaller than the threshold δ1, step 708 is performed immediately without adjusting the chromatic quantization coefficient. If QPb is greater than threshold δ1, step 705 is performed (704). The reason for such a condition is that the occurrence of color difference decreases with higher image quality and becomes worse with lower image quality. Therefore, delta 1 becomes a threshold of high quality and low quality and QPb is a measure of the quality of the current image. δ1 may vary depending on the performance of the quantization method and other coding methods, and may be determined experimentally.

If the image quality of the current image is low, that is, QPb is larger than the threshold δ1, the absolute value of d1 is compared with the predetermined threshold δ2 (705). If d1 is smaller than δ2, it is determined that the color difference phenomenon is not severe, and step 708 is immediately performed without adjusting the chromatic quantization coefficient. This is because when the d1 is small, that is, when the difference in the quantization coefficients between regions is small, the color difference phenomenon is less likely to occur. Here, the value of δ2 is closely related to the image quality. As the quality is lower, that is, the larger the value of QPb, the color difference is more severe. In this case, the color difference is easily generated even if d1 is small. Therefore, d1 becomes smaller as QPb becomes larger. The exact value of δ 2, like δ 1, may vary depending on the performance of the quantization method and other coding methods, and may be determined experimentally.

If both of the steps 704 and 705 satisfy the conditions, a color difference phenomenon occurs. Therefore, it is necessary to limit the change in the quantization coefficient for the chromaticity component for each region. To this end, the quantization coefficient for the chromaticity component is changed to the value obtained by subtracting the quantization coefficient for the chromaticity component of the background region 405 by a predetermined value d2 (707). At this time, the absolute value of d2 is smaller than the absolute value of d1 so that the difference in the quantization coefficient for the chromaticity component between the region of interest 404 and the background region 405 is small (706). If the color difference is severe, set d2 to 0 to reduce the color difference as much as possible.

Using the quantization coefficients of the luminance components of the ROI 404 and the background region 405 obtained in operation 702, the two luminance components are included in the block belonging to the variation region 403 to reduce the boundary effect on the luminance components. The coefficient quantized by the value between the quantized coefficients is used for encoding and decoding (708). For the chromaticity component, the chromaticity quantization coefficient in the variation region 403 is determined using the quantization coefficients of the chromaticity components of the ROI 404 and the background region 405 obtained in operation 702 or 707 (operation 709).

8 shows some examples of how to change the values of luminance and chroma quantization coefficients in the variation region 403. First, in FIG. 8A, the variation region 403 is set to be the same for the luminance and the chromaticity, and the value of the chromatic quantization coefficient in the ROI 404 is set to be larger than the luminance quantization coefficient. The amount of change in chroma quantization coefficient between 405 is made smaller than the amount of change in luminance quantization coefficient. Here, the luminance and chromatic quantization coefficients, QPbLum and QPbChr, of the background region 405 are calculated using QPb as shown in FIG. Calculate using On the contrary, the same effect can be achieved by setting the value of the chromatic quantization coefficient in the region of interest 404 equal to the value of the luminance quantization coefficient and setting the value of the chromaticity quantization coefficient in the background region 405 smaller than the value of the luminance quantization coefficient. Can be obtained.

In FIG. 8B, the variation region 403 of the chrominance component of the chroma component is larger than the variation region 403 of the quantization coefficient of the luminance component, so that the amount of change in the quantization coefficient of the chroma component is smaller than the luminance component. In FIG. 8B, the boundary c of the background region 405 is the same in the case of the luminance component and the chroma component, but the boundary of the region of interest 404 is b in the luminance component and a in the chroma component. It can be seen that the larger the component, the smaller the amount of change in the quantization coefficient of the chromatic component. On the contrary, the boundary of the region of interest 404 is the same in the case of the luminance component and the chromaticity component, and the boundary of the background region 405 is changed so that the variation region 403 becomes larger in the chromaticity component so that the quantization coefficient variation of the chromaticity component is changed. It can also be made smaller.

8C and 8D, only the quantization coefficients of the luminance component are changed and the quantization coefficients of the chromaticity component use the same value in all regions. In FIG. 8C, the quantization coefficient of the chromaticity component is used for the background region 405 in all regions, and in FIG. 8D, the value of the region of interest 404 is used.

9A and 9B compare images obtained by encoding and decoding an actual image by a conventional method and results of encoding and decoding by applying the quantization method according to the present invention. 9A and 9B, the image on the left is a case where a conventional method is applied and the image on the right is a case where the method according to the present invention is applied. In FIG. 9A, the face part of the person is set as the ROI. In the image on the left side, the color of the human head is changed and spread by using different quantization coefficients in the ROI and the background area. On the other hand, in the image on the right side, the color difference quantization coefficient is used in the ROI and the background region using the same method as in FIG. In addition, in the left image of FIG. 9B, the color of the red light is spread on the white wavy portion in the center left portion of the screen. In the image on the right side, it is understood that such a color difference is removed using the same method as in FIG. 8C. have. Here, the phenomenon of color difference occurs because different quantization coefficients are used in each region, and the phenomenon of color difference spreading is performed by predictive encoding in time when video is encoded. That is, the color difference generated in one chapter is continuously spread on the image after being encoded with reference to the image.

As described above, in order to enhance encoding efficiency, when a single image is divided into regions and encoded, the subjective image quality can be improved by reducing the color difference for each region so that the amount of change in chromaticity quantization coefficient values of each region does not change significantly. This method is equally applicable to changing the quantization coefficient to control the bit rate for each unit area such as a macroblock.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

The best embodiments have been disclosed in the drawings and specification above. Although specific terms have been used herein, they are used only for the purpose of describing the present invention and are not used to limit the scope of the present invention as defined in the meaning or claims. Therefore, those skilled in the art will understand that various modifications and equivalent other embodiments are possible from this. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

1A is a block diagram schematically illustrating an embodiment of an image encoding apparatus through color quantization in consideration of a visual characteristic of a person according to the present invention.

1B is a block diagram schematically illustrating an embodiment of an image decoding apparatus through color quantization in consideration of a visual characteristic of a person according to the present invention.

FIG. 2A is a flowchart illustrating an embodiment of an image encoding method through color quantization in consideration of visual characteristics of a person performed in the apparatus of FIG. 1A.

FIG. 2B is a flowchart illustrating an embodiment of an image decoding method through color quantization considering a visual characteristic of a person performed in the apparatus of FIG. 1B.

3 is a diagram illustrating a configuration of a macroblock used as a basic unit in image compression.

4A and 4B are diagrams illustrating an exemplary embodiment in which a region is set for dividing and encoding a single image into slice structures.

5A and 5B illustrate an embodiment of a slice header.

6 is a diagram illustrating a relationship between a quantization coefficient QP, a luminance quantization coefficient, and a chromatic quantization coefficient.

7 is a flowchart illustrating a method of obtaining quantization coefficients for luminance and chromatic components in a region of interest, a transition region, and a background region.

8A to 8D illustrate a method of obtaining quantization coefficients for luminance and chromatic components in a region of interest, a transition region, and a background region.

9A and 9B are diagrams comparing images according to a conventional method and an encoding and decoding method according to the present invention in an actual image.

Claims (3)

Decoding quantization information and additional quantization information of the magnetic region included in one slice unit header from the bit string; Determining first quantization coefficients of macroblocks included in a first region by using the decoded quantization information and additional quantization information of the magnetic region; Obtaining second quantization coefficients of macroblocks included in a second region from the decoded quantization information of the magnetic region; And Performing inverse quantization of the macroblocks included in the first region using the determined first quantization coefficients, and performing inverse quantization of the macroblocks included in the second region using the second quantization coefficients. Image dequantization method comprising. The image inverse quantization method of claim 1, wherein the additional quantization information is quantization information of a region adjacent to the magnetic region. The image inverse quantization method of claim 2, wherein the region adjacent to the magnetic region is a region corresponding to a boundary of the magnetic region.

Images