VVC coding block fast partitioning method based on second-order difference

Abstract

The application discloses a kind of based on second-order difference VVC coding block fast division method, specific steps include: step 1, screening target size coding unit (CU), wherein target size is less than or equal to 32X32;Step 2, calculate the texture intensity in the horizontal and vertical direction of coding unit;Step 3, the ratio of maximum texture intensity in horizontal and vertical direction is compared with threshold β, and then the more suitable division mode and direction of coding unit is decided in advance.The fast division method of coding unit in VVC coding standard based on second-order difference proposed in the application reduces the overall coding complexity while ensuring coding efficiency and video quality.

Description

Technical Field

[0001] This invention belongs to the field of video encoding and decoding, specifically a fast VVC coding block partitioning method based on second-order difference. Background Technology

[0002] With the development of industries such as VR simulation video, 360-degree video, and real-time video, and the expansion from two-dimensional planes to multi-dimensional virtual video, and from standard dynamic range to high dynamic range, video encoding and decoding technologies are also constantly evolving. In 2020, JVET launched the next-generation multi-functional video coding standard H.266 / VVC (VVC for short), based on... The delta bitrate metric mechanism reduces the bitrate by an average of 41% compared to the previous generation standard. This is because the new generation of video coding standards, while retaining the previous block-based hybrid coding framework, introduces many efficient encoding and decoding tools. In terms of VVC coding block partitioning modes, in addition to retaining the original quadtree partitioning (QT), four additional partitioning modes have been added: horizontal binary tree partitioning (BH), vertical binary tree partitioning (BV), horizontal ternary tree partitioning (TH), and vertical ternary tree partitioning (TV).

[0003] In VVC, the optimal partitioning mode is selected from these five partitioning modes through exhaustive recursion. For example... Figure 1 The diagram illustrates the actual partitioning of coding units (CUs). The partitioning of each coding unit is related to its texture. For flat regions or regions with obvious texture directionality, there is a clear bias in the selection of partitioning modes, and attempting obviously unsuitable partitioning modes is redundant. This exhaustive partitioning process greatly increases the overall coding complexity. Therefore, it is necessary to take certain measures to reduce the complexity of the CU partitioning mode selection process. The previous generation of video coding standards only had one QT partitioning mode, which was a binary classification problem of whether to partition. For CUs with complex texture content, it tended to partition, and vice versa; in this case, it was only necessary to judge the texture complexity of the CU. However, the four new partitioning modes added in VVC are all related to direction. At this time, it is necessary to consider the texture direction of the CU to filter the candidate partitioning mode list. At the same time, there are three types of sub-blocks for these five partitioning modes. The more sub-blocks, the more complex the texture of the current CU. Therefore, the partitioning problem of coding units in VVC is more complex and needs to be considered from multiple perspectives.

[0004] Traditional image processing methods cannot be universally applied to CU partitioning, so it is necessary to improve and innovate the corresponding methods to better achieve a balance between coding efficiency and time saving. Summary of the Invention

[0005] The object of the present invention is to propose a fast VVC coding block partitioning method based on second-order difference, which reduces the complexity of the coding unit partitioning mode selection and saves the overall coding time while ensuring the video quality and coding efficiency.

[0006] The technical solution for implementing the present invention is as follows: In the first aspect, the present invention provides a fast VVC coding block partitioning method based on second-order difference, including the following steps:

[0007] Step 1: Obtain the width and height dimensions of the current coding unit, and perform the following operations on coding units smaller than or equal to 32X32.

[0008] Step 2: Calculate the texture intensities in the horizontal and vertical directions of the coding unit, and use the maximum texture intensity to represent the texture intensity in that direction.

[0009] Step 3: Calculate the threshold β according to the quantization parameter, compare the ratio of the maximum texture intensities in the two directions in Step 2 with β, and then screen the partitioning mode candidate list.

[0010] Further, for the calculation of the texture intensity, it is specifically as follows:

[0011] For a coding unit with a width of W and a height of H, the texture intensity between the jth, j + 2nd, and j + 4th columns in the horizontal direction is calculated using second-order difference, and the formula is as follows:

[0012]

[0013] where (i, j) represents the coordinates of the corresponding pixel point, and P(i, j) represents the pixel value at the corresponding coordinates (i, j); as shown in the formula, instead of using the second-order difference between adjacent columns, downsampling with a step size of 2 is performed on the column numbers; the texture intensities of all such CUs in the horizontal direction are calculated as follows:

[0014]

[0015] where j satisfies the condition: j ∈ (1 <= j < W) && (j % 4 == 1); that is, downsampling with a step size of 4 is performed on the width, and the first and last columns of adjacent two texture intensities coincide; the maximum texture intensity THmax in the horizontal direction is the maximum value among them;

[0016] The calculation formula in the vertical direction is as follows:

[0017] ]

[0018] where i satisfies the condition: i ∈ (1 <= i < H) && (i % 4 == 1); the maximum texture intensity TVmax in the vertical direction is the maximum value among them.

[0019] Furthermore, the rules for candidate pattern selection are as follows:

[0020] (1) When THmax / TVmax>β, it means that the texture difference in the horizontal direction of the current CU is more obvious, and the coding block is more inclined to the vertical direction partitioning mode. Therefore, the horizontal partitioning mode is skipped; for CU with width < height, only the horizontal ternary tree partitioning is skipped.

[0021] (2) When TVmax / THmax>β, it means that the texture difference in the vertical direction of the current CU is more obvious, and the coding block is more inclined to the horizontal division mode, so the vertical division mode is skipped; for CUs with width>height, only the vertical ternary tree division is skipped.

[0022] (3) When neither (1) nor (2) is satisfied, it means that the texture direction difference of the current CU is not obvious, so it is more suitable for binary tree partitioning with weaker directionality, and therefore skips ternary tree partitioning.

[0023] Furthermore, the threshold β in step 3 is related to the quantization parameter QP, and the value of β is as follows:

[0024]

[0025] QP represents the quantization parameter.

[0026] Furthermore, for encoding units that do not meet the size or threshold conditions, the original partitioning process is directly carried out, that is, the optimal partitioning mode is selected by exhaustive search.

[0027] Furthermore, in step 3, after removing unsuitable patterns from the candidate list, the remaining candidate patterns will be processed according to the original partitioning steps.

[0028] In a second aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described in the first aspect.

[0029] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0030] Fourthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0031] Compared with existing algorithms, this method has the following advantages:

[0032] (1) Compared with the common method of calculating the texture intensity in two directions respectively after extracting the directional gradient for edge detection, the proposed method simplifies it to one step, directly using the similarity between adjacent pixels to represent the texture intensity for comparison;

[0033] (2) Meanwhile, the candidate mode list is screened from two aspects of the division direction and the complexity of the division mode, and the CUs with different aspect ratios are adaptively processed. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the actual division of the coding block.

[0035] Figure 2 It is a flowchart of the fast division algorithm for the VVC coding block based on the second-order difference. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] The present invention uses the texture intensity in the horizontal and vertical directions of the coding unit to screen the candidate list of division modes. Combining Figure 2 , the specific steps are as follows:

[0037] Step 1, obtain the width and height dimensions of the current coding unit, and perform the following operations for coding units less than or equal to 32X32;

[0038] Step 2, calculate the texture intensity in the horizontal and vertical directions of the coding unit, and use the maximum texture intensity to represent the texture intensity in that direction;

[0039] Step 2.1 Calculate all the horizontal texture intensities:

[0040]

[0041] where j satisfies the condition: j ∈ (1 <= j < W) && (j % 4 == 1). The maximum horizontal texture intensity THmax is the maximum value of all the horizontal texture intensities.

[0042] Step 2.2 Calculate all the vertical texture intensities:

[0043]

[0044] where i satisfies the condition: i ∈ (1 <= i < H) && (i % 4 == 1). The maximum vertical texture intensity TVmax is the maximum value of all the vertical texture intensities.

[0045] Step 3, compare the ratio of the maximum texture intensities in the two directions in Step 2 with the threshold β, and then screen the candidate list of division modes;

[0046] Step 3.1 Calculate the corresponding threshold β according to the following formula:

[0047]

[0048] Step 3.2 compare the THmax / TVmax values ​​with β and perform corresponding filtering:

[0049] (1) When THmax / TVmax>β, it means that the texture difference in the horizontal direction of the current CU is more obvious, and the coding block is more inclined to the vertical division mode, so the horizontal division mode is skipped; for CU with width < height, only the horizontal ternary tree division (TTH) is skipped.

[0050] (2) When TVmax / THmax>β, it means that the texture difference in the vertical direction of the current CU is more obvious, and the coding block is more inclined to the horizontal division mode, so the vertical division mode is skipped; for CUs with width>height, only the vertical ternary tree division (TTV) is skipped.

[0051] (3) When neither (1) nor (2) is satisfied, it means that the texture direction difference of the current CU is not obvious, so it is more suitable for binary tree partitioning (BT) partitioning with weaker directionality, and therefore skips ternary tree partitioning (TT) partitioning.

[0052] The technical solution of the present invention will be further described in detail below through embodiments.

[0053] Example

[0054] This embodiment demonstrates a fast VVC coded block partitioning method based on second-order difference, the process of which is as follows: Figure 2 As shown, before dividing the coding unit, it is first determined whether the coding unit meets the block size condition of the fast algorithm; for those that do not meet the condition, the original division process is directly carried out, and for those that meet the condition, the fast division algorithm is used.

[0055] For each coding unit that meets the conditions, the ratio of the maximum horizontal and vertical texture intensity is calculated. Then, the ratio is compared with the threshold β. Finally, the candidate list of partitioning modes is filtered based on the comparison results, leaving the partitioning modes that are more likely to be suitable.

[0056] The original recursive partitioning of the filtered pattern list is used to find the most suitable partitioning pattern. In this way, the number of candidate patterns for some coding units is adaptively reduced, thereby reducing the coding complexity.

[0057] The above method was compared with the original VTM10.0 coding performance. Table 1 shows the comparison of coding time, BDBR, and BD-PSNR coding performance. BDBR represents the change in bit rate compared to the original VTM10.0 coding method under the same objective video quality (i.e., PSNR). Coding time ΔT represents the time saving compared to the original VTM10.0 coding method. BD-PSNR represents the loss in video quality obtained by this method compared to the original VTM10.0 coding method under the same bit rate. The experimental results show that compared to the original encoder, this method can save overall coding time while maintaining reconstructed video quality and coding efficiency.

[0058] Table 1 Comparison of experimental results between the method of this invention and the VTM10.0 method

[0059]

[0060]

[0061] This invention is not limited to the contents involved in the claims and the above embodiments. Any invention created based on the concept of this invention should fall within the protection scope of this invention.

Claims

1. A fast VVC coding block partitioning method based on second-order difference, characterized in that, Includes the following steps: Step 1: Obtain the width and height dimensions of the current coding unit. For coding units that are less than or equal to 32x32, proceed to the next step. Step 2: Calculate the texture intensity in the horizontal and vertical directions of the coding unit, and use the maximum texture intensity to represent the texture intensity in that direction; the calculation of texture intensity is as follows: For a coding unit with width W and height H, the texture intensity between the j-th, j+2-th, and j+4-th columns in the horizontal direction is calculated using the second-order difference, as shown in the following formula: ； Where (i,j) represents the coordinates of the corresponding pixel point, and P(i,j) represents the pixel value at coordinates (i,j); the calculation of all texture intensities in the horizontal direction of this size CU is as follows: ； Where j satisfies the following condition: ; That is, downsampling with a step size of 4 is performed on the width, and the first and last columns of two adjacent texture intensities overlap; the maximum texture intensity THmax in the horizontal direction is the maximum value among them. The calculation formula in the vertical direction is as follows: ； Where i satisfies the following condition: The maximum texture intensity TVmax in the vertical direction is the largest of these values. Step 3: Calculate the threshold β based on the quantization parameters, compare the maximum texture intensity ratio in the two directions in Step 2 with β, and then filter the candidate list of partitioning modes.

2. The fast VVC coding block partitioning method based on second-order difference according to claim 1, characterized in that, The specific rules for candidate pattern selection are as follows: (1) When THmax / TVmax>β, it means that the texture difference in the horizontal direction of the current CU is more obvious, and the horizontal partitioning mode is skipped; for CUs with width < height, only the horizontal ternary tree partitioning is skipped. (2) When TVmax / THmax>β, it means that the texture difference in the vertical direction of the current CU is more obvious, and the vertical partitioning mode is skipped; for CUs with width>height, only the vertical ternary tree partitioning is skipped. (3) When neither (1) nor (2) is satisfied, it means that the texture direction difference of the current CU is not obvious, and the ternary tree partitioning is skipped.

3. The fast VVC coding block partitioning method based on second-order difference according to claim 2, characterized in that, The threshold β in step 3 is related to the quantization parameter QP, and the values ​​of β are as follows: ； QP represents the quantization parameter.

4. The fast VVC coding block partitioning method based on second-order difference according to claim 1, characterized in that, For coding units that do not meet the size or threshold conditions, the optimal partitioning mode is selected by exhaustive search.

5. The fast VVC coding block partitioning method based on second-order difference according to claim 1, characterized in that, For step 3, after removing unsuitable patterns from the candidate list, the remaining candidate patterns are used to select the best partitioning pattern through exhaustive search.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1-5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-5.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-5.

Images