Coding method, system and device for AVS3 intra-frame angle prediction
By recalculating and recoding the reference pixel address, the problem of excessive intra-angle prediction hardware resources in the prior art is solved, and more efficient hardware resource usage and dynamic power consumption are achieved.
Patent Information
- Application Number
- CN202410880642.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-02
- Publication Date
- 2024-12-06
AI Technical Summary
In the existing encoding technology, there are too many hardware resources required for intra-angle prediction, which is difficult to apply in actual systems.
By recalculating and recoding the reference pixel address of each predicted pixel, the use of hardware resources is simplified, and the encoding method of AVS3 intra-angle prediction is adopted, including obtaining coordinates based on the number of modes, calculating the intermediate amount of the reference pixel address and filtering parameters, and performing Reference pixel queue offset and recoding to reduce the use of hardware resources.
It reduces the occupation of hardware resources, improves the utilization rate of hardware resources, simplifies the computing process, and reduces dynamic power consumption and logical area.
Smart Images

Figure CN119094780A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of image and video data compression processing, in particular to an encoding method, system and device for AVS3 intra-frame angle prediction. Background Art
[0002] The intra prediction method is divided into three parts: RMD, INTRAPRED, and RD-COST decisions. Although the prediction results are relatively accurate, it is necessary to traverse and calculate the prediction values of all types of CUs of 65 intra prediction modes, among which the most complex angle prediction modes have 60.
[0003] Taking the AVS3 encoder as an example, the process of intra-frame angle prediction calculation includes the following steps: 1. Obtain the reference pixels required by the current CU: the reference pixels are divided into left boundary pixels, upper boundary pixels, and upper left corner pixels. The number of left boundary pixels and upper boundary pixels is twice the height and width of the current CU, respectively. 2. Determine the interpolation filter parameters: According to the number of prediction modes required for the current CU, the interpolation filter parameters for determining the subsequent reference pixel address calculation are obtained by looking up the table. 3. Calculate the required reference pixel address: The reference pixel address is determined by the parameters obtained by looking up the table and the type of the current angle mode. The 4 addresses of the pixel pointed to by the current angle and its neighboring pixels are obtained. If the address exceeds 2 times the current side length, the address is limited to the boundary address. 4. Prediction pixel calculation: After obtaining the 4 reference pixel addresses, select the required pixels from the already obtained left boundary reference pixels, upper boundary reference pixels, and upper left corner reference pixels, and input the 4-tap filter calculated in step 2 to obtain the predicted pixel value. Since a single predicted pixel needs to rely on four reference pixels for calculation, multi-pixel parallel prediction is often required in real-time encoding systems, which requires more hardware resources to build. The traditional angle prediction hardware architecture requires extremely high hardware resources to implement and is difficult to apply in actual systems. Summary of the invention
[0004] The present invention aims to solve the problem of excessive prediction amount faced by angle prediction in existing coding technologies.
[0005] The present invention provides an encoding method, system and device for AVS3 intra-frame angle prediction.
[0006] An encoding method for AVS3 intra-frame angle prediction, comprising:
[0007] Step 1, according to the mode number of the current coding unit, the coordinates di vDxy or di vDyx are obtained, and the two numbers of the coordinates form a sequence {M, N}, and according to M, N and the horizontal coordinate or vertical coordinate of the predicted pixel to be calculated, the intermediate quantity and filtering parameter of the reference pixel address are calculated, and two or more reference pixel addresses required for the current predicted pixel position are obtained according to the intermediate quantity of the reference pixel address, and the reference pixel address with the smallest value is selected as the minimum reference pixel address;
[0008] Step 2: if the minimum reference pixel address is greater than or equal to a preset threshold, an offset operation is performed on the reference pixel queue so that the lowest-order reference pixel is located at address 0 in the offset reference pixel queue;
[0009] Step 3, cutting the reference pixel queue from the lowest bit, and re-encoding the offset reference pixel address so that the reference pixel address is within the cut reference pixel queue;
[0010] Step 4: Calculate the predicted value of the predicted pixel at the current position according to the filtering parameters obtained in step 1, the re-encoded reference pixel address obtained in step 3, and the reference pixel queue.
[0011] Preferably, in step 1, the intermediate quantity of the reference pixel address includes the intermediate quantity of the horizontal coordinate and the intermediate quantity of the vertical coordinate. If the pattern number is the coordinate di vDxy, the intermediate quantity of the horizontal coordinate is calculated; if the pattern number is the coordinate di vDyx, the intermediate quantity of the vertical coordinate is calculated.
[0012] Calculating the intermediate amount of the horizontal coordinate includes the following steps: setting the horizontal coordinate of the predicted pixel to be calculated to i; multiplying M by i+1, and then shifting the overall result right by N bits;
[0013] Calculating the intermediate quantity of the ordinate includes the following steps: setting the ordinate of the predicted pixel to be calculated to i; multiplying M by i+1, and then shifting the overall result right by N bits.
[0014] Preferably, in step 1, the filtering parameters include abscissa parameters and ordinate parameters. If the mode number is the coordinate di vDxy, the abscissa parameter is calculated; if the mode number is the coordinate di vDyx, the ordinate parameter is calculated.
[0015] Calculating the horizontal coordinate parameter includes the following steps: (1) obtaining the middle value of the reference pixel address corresponding to the current position, and shifting it right by 5 bits; (2) setting the horizontal coordinate of the predicted pixel to be calculated to i; (3) multiplying i+1 by M, shifting it left by 5 bits, and then shifting the overall result right by N bits; (4) subtracting the number obtained in (3) from the number obtained in (1);
[0016] Calculating the vertical coordinate parameter includes the following steps: (1) obtaining the middle value of the reference pixel address corresponding to the current position and shifting it right by 5 bits; (2) setting the vertical coordinate of the predicted pixel to be calculated to i; (3) multiplying i+1 by M, shifting it left by 5 bits, and then shifting the overall result right by N bits; (4) subtracting the number obtained in (3) from the number obtained in (1).
[0017] Preferably, in step 4, the corresponding pixel value is obtained according to the re-encoded reference pixel address, and the pixel value is multiplied with the preset value B-filter parameter, the preset value C-filter parameter, the preset value B+filter parameter and the filter parameter respectively, the product is added to 64 and then right-shifted 7 bits to obtain the predicted value of the predicted pixel at the current position.
[0018] Preferably, the re-encoded reference pixel address is: the offset reference pixel address-the minimum reference pixel address+the preset threshold.
[0019] Preferably, in step 2, the reference pixel queue is shifted right, and the number of pixels shifted right is the difference between the minimum reference pixel address and a preset threshold.
[0020] Preferably, in step 1, the mode number range is 0-65. If the current mode number is in the interval [13, 23], the reference pixel addresses are: i+context-1, i+context, i+context+1, i+context+2; otherwise, they are: i+context+1, i+context, i+context-1, i+context-2; where context is the intermediate value of the reference pixel address.
[0021] Preferably, in step 1, a parameter i Dxy is defined, which is 1 when the number of prediction modes is in the interval (12, 24), otherwise -1; a parameter ui xyf l ag is defined, which is 0 when the number of prediction modes is less than 25, otherwise 1; and the reference pixel address includes the following calculation steps:
[0022] (1) If the mode number is the coordinate d ivDxy, then calculate the parameter iXx; if the mode number is the coordinate d ivDyx, then calculate the parameter iYy; iXx, its value is the difference between i and the middle value of the horizontal coordinate; iYy, its value is the difference between i and the middle value of the vertical coordinate;
[0023] (2) When iDxy<0 and uixyf l ag is equal to 0, obtain one of the four reference pixel addresses iX, whose value is i + the middle value of the horizontal coordinate; when i Dxy<0 and uixyf l ag is not equal to 0, obtain one of the four reference pixel addresses iY, whose value is i + the middle value of the vertical coordinate; when i Dxy>0, determine whether iYy is less than 0. If iYy is less than 0, obtain one of the four reference pixel addresses iX, whose value is iXx obtained in (1); if iYy is greater than 0, obtain one of the four reference pixel addresses iY, whose value is iYy obtained in (1);
[0024] (3) To obtain the remaining three reference pixel addresses, first determine whether iDxy is less than 0. If i Dxy is less than 0, when the reference pixel address obtained in (2) is iX, the remaining three reference pixel addresses are iX-1, iX+1 and iX+2 respectively; when the reference pixel address obtained in (2) is iY, the remaining three reference pixel addresses are iY-1, iY+1 and iY+2 respectively; if i Dxy is greater than 0, when the reference pixel address obtained in (2) is iX, the remaining three reference pixel addresses are iX+1, iX-1 and iX-2 respectively; when the reference pixel address obtained in (2) is iY, the remaining three reference pixel addresses are iY+1, iY-1 and iY-2 respectively;
[0025] (4) When the reference pixel address obtained in (2) is iX, the address obtained based on iX is compared with twice the width of the coding unit. If it is greater than twice the width, the address is limited to twice the width. When the reference pixel address obtained in (2) is iY, the address obtained based on iY is compared with twice the height of the coding unit. If it is greater than twice the height, the address is limited to twice the height.
[0026] A coding system for intra-frame angle prediction of AVS3 is used to implement the coding method for intra-frame angle prediction of AVS3, comprising: a reference pixel queue offset module, an offset calculation module and a re-encoding module, wherein the reference pixel queue is offset according to the reference pixel address, the offset calculation module calculates the pixel queue offset and outputs the reference pixel queue offset, and the re-encoding module re-encodes the reference pixel address and outputs the re-encoded reference pixel address.
[0027] An encoding device for angle prediction within an AVS3 frame is used to implement the encoding method for angle prediction within an AVS3 frame, including a hardware circuit. The hardware circuit is designed as an ASIC and circuit synthesis is performed under the clock constraint of a 28nm process and a maximum main frequency of 500Mhz.
[0028] Compared with the prior art, the present invention has the following advantages and effects:
[0029] The present invention simplifies the 8-bit 129-to-1 MUX into an 8-bit 7-to-1 MUX by re-encoding and decoding the four reference pixel addresses required for each predicted pixel and self-encoding and decoding the reference pixel queue, thereby reducing the use of multiple large-bit-width selectors that occupy extremely high hardware resources, simplifying the calculation process, and improving the utilization rate of hardware resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0031] Figure 1 It is a flow chart of the present invention.
[0032] Figure 2 A schematic diagram of reference pixel queue offset in the encoding method of the present invention.
[0033] Figure 3 It is a schematic diagram of the hardware circuit of the present invention.
[0034] Figure 4 The figure is a comparison diagram of the area of the circuit of the present invention and the existing circuit when the pixel parallelism is 8, 16, 32 and 64.
[0035] Figure 5 The figure is a comparison diagram of dynamic power consumption of the circuit of the present invention and the existing circuit when the pixel parallelism is 8, 16, 32 and 64. DETAILED DESCRIPTION
[0036] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0037] like Figure 1-5 As shown, a coding method for intra-frame angle prediction of AVS3 includes:
[0038] Step 1, according to the mode number of the current coding unit, the coordinates d ivDxy or d ivDyx are obtained, and the two numbers of the coordinates form a sequence {M, N}, and according to M, N and the horizontal coordinate or vertical coordinate of the predicted pixel to be calculated, the intermediate quantity and filtering parameter of the reference pixel address are calculated, and two or more reference pixel addresses required for the current predicted pixel position are obtained according to the intermediate quantity of the reference pixel address, and the reference pixel address with the smallest value is selected from the reference pixel addresses as the minimum reference pixel address;
[0039] Step 2: if the minimum reference pixel address is greater than or equal to a preset threshold, an offset operation is performed on the reference pixel queue so that the lowest-order reference pixel is located at address 0 in the offset reference pixel queue;
[0040] Step 3, cutting the reference pixel queue from the lowest bit, and re-encoding the offset reference pixel address so that the reference pixel address is within the cut reference pixel queue;
[0041] Step 4: Calculate the predicted value of the predicted pixel at the current position according to the filtering parameters obtained in step 1, the re-encoded reference pixel address obtained in step 3, and the reference pixel queue.
[0042] In a specific implementation, in step 1, according to the provisions of the AVS3 standard, the intra-frame prediction mode range is 0-32 (if the intra-frame angle prediction extension is turned on, the range is extended to 0-65, that is, the mode number range is 0-65), where except for modes 0, 1, 2, 12, and 24, all other modes are angle prediction. The reference pixel required for each pixel is related to the position of its pixel. For example, in an 8x8 block, the reference pixel required for the position with coordinates (0,0) and (0,1) is different. The width and height are related to the rules of intra-frame prediction. The queue of reference pixels is limited to 2 times the width + 2 times the height. For example, when the lower left pixel of the 8x8 block is (7,7), the mode is 32 (pointing to the upper right), the calculated address will be greater than 2x8=16 (possibly 20), and the address needs to be limited to 16. Therefore, after determining to start the intra-frame angle prediction program, it is necessary to input the width and height of the current coding unit, the mode number, and the offset coordinates of the required prediction pixel based on the upper left corner pixel of the current coding unit. When the mode number is between 13 and 23, the direction of the intra-frame angle prediction is the upper right corner (between 0 and 90 degrees). At this time, 4 reference pixel addresses may span the middle of the queue, the upper left corner pixel (that is, there may be both the upper side, the left side, and the upper left corner). If it is outside this interval, the reference pixel address only refers to the reference pixel on a single side (only the left side is referenced, then all addresses only refer to the upper side).
[0043] In a specific implementation, in step 1, firstly, coordinates divDxy or divDyx are obtained by looking up the table according to the mode number, and the two numbers of the coordinates are composed of {imult, ishift}. The intermediate quantity of the reference pixel address includes the intermediate quantity of the horizontal coordinate and the intermediate quantity of the vertical coordinate. If the mode number is the coordinate divDxy, the intermediate quantity of the horizontal coordinate is calculated, and if the mode number is the coordinate divDyx, the intermediate quantity of the vertical coordinate is calculated; calculating the intermediate quantity of the horizontal coordinate includes the following steps: setting the horizontal coordinate of the predicted pixel to be calculated as i; multiplying M by i+1, and then shifting the overall result right by N bits; calculating the intermediate quantity of the vertical coordinate includes the following steps: setting the vertical coordinate of the predicted pixel to be calculated as i; multiplying M by i+1, and then shifting the overall result right by N bits. An example of the calculation formula of the reference pixel address context: context = ((i+1)*imult<<5)>>is shift, where i represents the horizontal coordinate or vertical coordinate of the predicted pixel to be calculated. If the current mode number is in the interval [13,23], the value of i is the horizontal coordinate, otherwise it is the vertical coordinate.
[0044] In a specific implementation, in step 1, firstly, a table is looked up according to the mode number to obtain the coordinate divDxy or divDyx, where the two numbers of the coordinate are composed of {imult,ishift}. The filtering parameters include a horizontal coordinate parameter and a vertical coordinate parameter. If the mode number is obtained as the coordinate divDxy, the horizontal coordinate parameter is calculated. If the mode number is obtained as the coordinate d ivDyx, the vertical coordinate parameter is calculated. The calculation of the horizontal coordinate parameter includes the following steps: (1) obtaining the middle value of the reference pixel address corresponding to the current position and shifting it right by 5 bits; (2) setting the horizontal coordinate of the predicted pixel to be calculated as i; (3) multiplying i+1 by M, shifting it left by 5 bits, and then shifting the overall result right by N bits; (4) subtracting the number obtained in (3) from the number obtained in (1); the calculation of the vertical coordinate parameter includes the following steps: (1) obtaining the middle value of the reference pixel address corresponding to the current position and shifting it right by 5 bits; (2) setting the vertical coordinate of the predicted pixel to be calculated as i; (3) multiplying i+1 by M, shifting it left by 5 bits, and then shifting the overall result right by N bits; (4) subtracting the number obtained in (3) from the number obtained in (1). The calculation formula of the filter parameter offset is as follows: offset = (((i+1)*imult<<5)>>ishift)-(((i+1)*imult>>ishift)<<5), where i represents the horizontal coordinate or vertical coordinate of the predicted pixel to be calculated. If the current mode number is in the interval [13,23], the value of i is the horizontal coordinate, otherwise it is the vertical coordinate.
[0045] In a specific implementation, in step 1, the calculation of the intra-frame prediction pixel in the existing AVS3 standard is obtained by calculating 4 reference pixels. When the calculation of the intra-frame prediction pixel changes, the corresponding number of reference pixels can be selected according to the change. If the current mode number is in the interval [13,23], the reference pixel addresses are: i+context-1, i+context, i+context+1, i+context+2; otherwise, they are: i+context+1, i+context, i+context-1, i+context-2; where context is the middle value of the reference pixel address. The specific operations are as follows:
[0046] Define parameter iDxy, when the number of prediction modes is in the interval (12, 24), it is 1, otherwise it is -1; define parameter uixyf l ag, when the number of prediction modes is less than 25, it is 0, otherwise it is 1; the reference pixel address includes the following calculation steps:
[0047] (1) If the mode number is the coordinate d ivDxy, then calculate the parameter iXx; if the mode number is the coordinate d ivDyx, then calculate the parameter iYy; iXx, its value is the difference between i and the middle value of the horizontal coordinate; iYy, its value is the difference between i and the middle value of the vertical coordinate;
[0048] (2) When iDxy<0 and uixyf l ag is equal to 0, obtain one of the four reference pixel addresses iX, whose value is i + the middle value of the horizontal coordinate; when i Dxy<0 and uixyf l ag is not equal to 0, obtain one of the four reference pixel addresses iY, whose value is i + the middle value of the vertical coordinate; when i Dxy>0, determine whether iYy is less than 0. If iYy is less than 0, obtain one of the four reference pixel addresses iX, whose value is iXx obtained in (1); if iYy is greater than 0, obtain one of the four reference pixel addresses iY, whose value is iYy obtained in (1);
[0049] (3) To obtain the remaining three reference pixel addresses, first determine whether iDxy is less than 0. If i Dxy is less than 0, when the reference pixel address obtained in (2) is iX, the remaining three reference pixel addresses are iX-1, iX+1 and iX+2 respectively; when the reference pixel address obtained in (2) is iY, the remaining three reference pixel addresses are iY-1, iY+1 and iY+2 respectively; if i Dxy is greater than 0, when the reference pixel address obtained in (2) is iX, the remaining three reference pixel addresses are iX+1, iX-1 and iX-2 respectively; when the reference pixel address obtained in (2) is iY, the remaining three reference pixel addresses are iY+1, iY-1 and iY-2 respectively;
[0050] (4) When the reference pixel address obtained in (2) is iX, the address obtained based on iX is compared with twice the width of the coding unit. If it is greater than twice the width, the address is limited to twice the width. When the reference pixel address obtained in (2) is iY, the address obtained based on iY is compared with twice the height of the coding unit. If it is greater than twice the height, the address is limited to twice the height.
[0051] In a specific implementation, in step 2, the preset threshold is 3, and the reference pixel queue is judged whether an offset operation is required based on the comparison between the minimum reference pixel address and 3. If the minimum reference pixel address is less than 3, the reference pixel queue is not offset. Otherwise, the reference pixel is right-shifted, and the number of pixels shifted right is the difference between the minimum reference pixel address and 3, so that the lowest-order reference pixel is located at the address 0 in the reference pixel queue after the offset.
[0052] In a specific embodiment, in step 3, the reference pixel queue is cropped from the lowest bit, and the length after cropping is changed from the original 129 pixels to 7 pixels. Addr[0] is the smallest reference pixel address obtained in the above step. It is subtracted from 3. If the result is a negative number, the reference pixel queue offset is 0, otherwise it is the result of the difference. The reference pixel address is re-encoded based on the reference pixel queue offset. If the offset is not equal to 0, the 4 re-encoded reference pixel addresses addr[0], addr[1], addr[2], and addr[3] are: 3, addr[1]-addr[0]+3, addr[2]-addr[0]+3, addr[3]-addr[0]+3, and the re-encoding formula example is as follows:
[0053] pixel_offset=addr[0]-3<0?0:addr[0]-3
[0054] addr[n]=pixel_offset! =0? addr[n]-addr[0]+3:addr[n].
[0055] Where pixel_offset is the reference pixel queue offset. addr[n] represents the re-encoded reference pixel address, and n is in the range of [0,3], representing the 4 reference pixels required for each predicted pixel calculation.
[0056] In a specific implementation, in step 4, the corresponding pixel value is obtained according to the re-encoded reference pixel address, and the pixel value is multiplied by the preset value B-filter parameter, the preset value C-filter parameter, the preset value B+filter parameter and the filter parameter respectively, and the product is added to 64 and right-shifted by 7 bits to obtain the predicted value of the predicted pixel at the current position. The formula example is as follows:
[0057] pred_pixel=(pixel_array_7[addr0]*(32-offset)
[0058] +pixel_array_7[addr1]*(64-offset)
[0059] +pixel_array_7[addr2]*(32+offset)
[0060] +pixel_array_7[addr3]*offset+64)>>7;
[0061] Where pred_pixel is the predicted pixel value, pixel_array_7 is the lowest 7-pixel reference pixel array intercepted after offset, addr0-addr3 are the re-encoded reference pixel addresses, and offset is the filtering parameter calculated according to the angle mode number in step 2.
[0062] An encoding system for intra-frame angle prediction of AVS3 includes: a reference pixel queue offset module, an offset calculation module and a re-encoding module. The reference pixel queue is offset according to the reference pixel address. The offset calculation module calculates the pixel queue offset and outputs the reference pixel queue offset. The re-encoding module re-encodes the reference pixel address and outputs the re-encoded reference pixel address.
[0063] An encoding device for intra-frame angle prediction of AVS3 includes a hardware circuit, which is designed by ASIC and described by SystemVer il og language. Synopsys Design Compiler is used to perform circuit synthesis under the clock constraint of 28nm process and maximum main frequency of 500Mhz. When 8 pixels are parallelly calculated, the angle prediction circuit area implemented by the method of the present invention is 0.014mm2 and the dynamic power consumption is 1.9617mW; the circuit area implemented by the traditional angle prediction method is 0.019mm2 and the dynamic power consumption is 2.2199mW, which saves 26% of logic usage and 11.6% of dynamic power consumption.
[0064] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A coding method for intra-frame angle prediction of AVS3, characterized in that: include: Step 1, according to the mode number of the current coding unit, the coordinates divDxy or divDyx are obtained, and the two numbers of the coordinates form a sequence {M, N}. According to M, N and the horizontal coordinate or vertical coordinate of the predicted pixel to be calculated, the intermediate quantity and filtering parameters of the reference pixel address are calculated, and two or more reference pixel addresses required for the current predicted pixel position are obtained according to the intermediate quantity of the reference pixel address, and the reference pixel address with the smallest value is selected as the minimum reference pixel address; Step 2: if the minimum reference pixel address is greater than or equal to a preset threshold, an offset operation is performed on the reference pixel queue so that the lowest-order reference pixel is located at address 0 in the offset reference pixel queue; Step 3, cutting the reference pixel queue from the lowest bit, and re-encoding the offset reference pixel address so that the reference pixel address is within the cut reference pixel queue; Step 4: Calculate the predicted value of the predicted pixel at the current position according to the filtering parameters obtained in step 1, the re-encoded reference pixel address obtained in step 3, and the reference pixel queue.
2. The encoding method for AVS3 intra-frame angle prediction according to claim 1, characterized in that: In the step 1, the intermediate quantity of the reference pixel address includes the intermediate quantity of the horizontal coordinate and the intermediate quantity of the vertical coordinate. If the mode number is the coordinate divDxy, the intermediate quantity of the horizontal coordinate is calculated, and if the mode number is the coordinate divDyx, the intermediate quantity of the vertical coordinate is calculated; Calculating the intermediate amount of the horizontal coordinate includes the following steps: setting the horizontal coordinate of the predicted pixel to be calculated to i; multiplying M by i+1, and then shifting the overall result right by N bits; Calculating the intermediate quantity of the ordinate includes the following steps: setting the ordinate of the predicted pixel to be calculated to i; multiplying M by i+1, and then shifting the overall result right by N bits.
3. The encoding method for AVS3 intra-frame angle prediction according to claim 1, characterized in that: In step 1, the filtering parameters include abscissa parameters and ordinate parameters. If the mode number is obtained as a coordinate divDxy, the abscissa parameter is calculated. If the mode number is obtained as a coordinate divDyx, the ordinate parameter is calculated. Calculating the horizontal coordinate parameter includes the following steps: (1) obtaining the middle value of the reference pixel address corresponding to the current position, and shifting it right by 5 bits; (2) setting the horizontal coordinate of the predicted pixel to be calculated to i; (3) multiplying i+1 by M, shifting it left by 5 bits, and then shifting the overall result right by N bits; (4) subtracting the number obtained in (3) from the number obtained in (1); Calculating the vertical coordinate parameter includes the following steps: (1) obtaining the middle value of the reference pixel address corresponding to the current position and shifting it right by 5 bits; (2) setting the vertical coordinate of the predicted pixel to be calculated to i; (3) multiplying i+1 by M, shifting it left by 5 bits, and then shifting the overall result right by N bits; (4) subtracting the number obtained in (3) from the number obtained in (1).
4. The encoding method for AVS3 intra-frame angle prediction according to claim 1, characterized in that: In step 4, the corresponding pixel value is obtained according to the re-encoded reference pixel address, and the pixel value is multiplied with the preset value B-filter parameter, the preset value C-filter parameter, the preset value B+filter parameter and the filter parameter respectively. The product is added to 64 and right-shifted 7 bits to obtain the predicted value of the predicted pixel at the current position.
5. The encoding method for AVS3 intra-frame angle prediction according to claim 1, characterized in that: The re-encoded reference pixel address is: the offset reference pixel address - the minimum reference pixel address + the preset threshold.
6. The coding method for AVS3 intra-frame angle prediction according to claim 1, characterized in that: In step 2, the reference pixel queue is shifted rightward, and the number of pixels shifted rightward is the difference between the minimum reference pixel address and the preset threshold.
7. The encoding method for AVS3 intra-frame angle prediction according to claim 2, characterized in that: In step 1, the mode number range is 0-65. If the current mode number is in the interval [13, 23], the reference pixel addresses are: i+context-1, i+context, i+context+1, i+context+2; otherwise, they are: i+context+1, i+context, i+context-1, i+context-2; where context is the intermediate value of the reference pixel address.
8. An AVS3 intra-frame angle prediction coding system, used to implement the AVS3 intra-frame angle prediction coding method according to any one of claims 1 to 7, characterized in that: It includes: a reference pixel queue offset module, an offset calculation module and a re-encoding module. The reference pixel queue is offset according to the reference pixel address. The offset calculation module calculates the pixel queue offset and outputs the reference pixel queue offset. The re-encoding module re-encodes the reference pixel address and outputs the re-encoded reference pixel address.
9. An encoding device for AVS3 intra-frame angle prediction, used to implement the encoding method for AVS3 intra-frame angle prediction according to any one of claims 1 to 7, characterized in that: The invention comprises a hardware circuit, wherein the hardware circuit is designed by ASIC and the circuit synthesis is performed under the clock constraint of 28nm process and maximum main frequency of 500Mhz.
Citation Information
Cited By
Predictive coding method and device and storage medium
CN119788877A
