WO2022022013A1 - Mode selection method and apparatus, computer-readable storage medium and electronic device - Google Patents

Abstract

The present disclosure provides a mode selection method and apparatus, a computer-readable storage medium and an electronic device, belonging to the technical field of video coding. The method comprises: selecting a number m of angle representative modes from angle modes provided by a preset coding standard, wherein m is smaller than the number of the angle modes provided by the preset coding standard; taking directions represented by the angle representative modes as alternative directions, and according to coding effect parameters of a current prediction unit in the various angle representative modes, selecting from the alternative directions a first direction corresponding to the current prediction unit; then selecting a candidate direction from the alternative directions according to the first direction; and according to the candidate direction, selecting a candidate mode from the modes provided by the preset coding standard. In this way, when multiple angle modes are screened, only the modes around the candidate direction, i.e., partial modes, are used as screening bases, thus during candidate mode selection, reducing the number of the modes needing to be determined wholly to some extent, and thereby reducing calculation amount and time consumption and improving efficiency.

Description

Mode selection method, apparatus, computer-readable storage medium, and electronic device

CROSS-REFERENCE TO RELATED APPLICATIONS

This disclosure claims the priority of a Chinese patent application with application number 202010761465.4 and entitled "Mode Selection Method, Apparatus, Computer-readable Storage Medium and Electronic Device" filed with the China Patent Office on July 31, 2020, the entire contents of which are provided by References are incorporated in this disclosure.

technical field

The present disclosure belongs to the technical field of video coding, and in particular, relates to a mode selection method, an apparatus, a computer-readable storage medium, and an electronic device.

Background technique

In order to improve the speed of video transmission, the video is often encoded by using a preset video encoding standard to reduce the amount of data to be transmitted. When using the preset coding standard for coding, the video image is often divided into coding tree units (coding tree units, CTUs) first, CTUs are divided into coding units (Coding units, CUs), and CUs are divided into prediction units (prediction, unit, PU). For each PU, it is necessary to perform traversal calculation on the current PU using each mode provided by the preset coding standard, and select candidate modes from the traversal calculation result. Then, a candidate list is generated according to the candidate mode, and the optimal mode corresponding to the coded PU adjacent to the current PU is also added to the candidate list. Finally, the current PU is traversed by using the candidate modes in the candidate list to select a mode with better coding effect for the current PU, for example, the optimal mode, and use this mode to complete the coding of the current PU.

In the process of selecting candidate modes, in the related art, all modes are often used as the screening basis, and the coding effect parameters of the current CU in each mode are calculated in turn, for example, the rate-distortion cost is calculated, and then according to the corresponding The encoding effect parameter filters candidate modes. In this way, when there are more PUs, or when the preset coding annotation provides multiple modes, for example, when the HEVC standard provides two non-angle modes and 33 angle modes, the traversal of each PU is performed. The way of calculating the encoding effect parameters in all modes will result in a large amount of calculation, long time consumption and low efficiency.

Overview

In view of this, the present disclosure provides a mode selection method, device, computer-readable storage medium and electronic device, which solve the problems of large amount of calculation, long time consumption and low efficiency when mode selection is performed to a certain extent .

In a first aspect, the present disclosure provides a mode selection method, which may include:

From the angle modes provided by the preset coding standard, select m angle representative modes; the m is less than the number of angle modes provided by the preset coding standard;

The direction represented by the angle representative mode is used as an alternative direction, and the first direction corresponding to the current PU is determined from the alternative directions according to the coding effect parameters of the current prediction unit PU in each of the angle representative modes; The encoding effect parameter in the angle representative mode corresponding to the first direction satisfies the first preset condition;

A candidate direction is selected from the candidate directions according to the first direction; the candidate directions include at least the first direction;

Based on the patterns around the candidate direction, a candidate pattern is selected from patterns provided by the preset coding standard.

In a second aspect, the present disclosure provides a mode selection apparatus, which may include:

a first selection mode, used for selecting m angle representative modes from the angle modes provided by the preset coding standard; the m is less than the number of angle modes provided by the preset coding standard;

a first determination mode, configured to use the direction represented by the angle representative mode as a candidate direction, and determine the current PU from the candidate directions according to the coding effect parameters of the current prediction unit PU in each of the angle representative modes The corresponding first direction; the angle corresponding to the first direction represents that the encoding effect parameter in the mode satisfies the first preset condition;

a second selection mode, configured to select a candidate direction from the candidate directions according to the first direction; the candidate directions include at least the first direction;

The third selection mode is used to select a candidate mode from modes provided by the preset coding standard based on modes around the candidate direction.

In a third aspect, the present disclosure provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the mode selection method according to the first aspect is implemented.

In a fourth aspect, the present disclosure provides an electronic device comprising: a processor, a memory, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program When implementing the mode selection method described in the first aspect.

For the related art, the present disclosure has the following advantages:

By selecting m angle representative modes from the angle modes provided by the preset coding standard, where m is less than the number of angle modes provided by the preset coding standard, the direction represented by the angle representative mode is taken as an alternative direction, and according to the current prediction unit PU For coding effect parameters in each angle representative mode, a first direction corresponding to the current PU is selected from the candidate directions, and then a candidate direction is selected from the candidate directions according to the first direction, and the candidate directions at least include the first direction. Finally, according to the candidate direction, a candidate mode is selected from the modes provided by the preset coding standard. In this way, the candidate direction in which the optimal mode may be located is determined by first selecting the angle representative mode, and the candidate mode is only selected in the mode around the candidate direction, that is, the direction with a small probability of having a mode with better coding effect is excluded, When screening for multiple angle modes, only the mode around the candidate direction, that is, the partial mode, is the basis for screening. In this way, the number of modes that need to be judged when screening the candidate mode can be reduced to a certain extent, thereby reducing the amount of calculation and consumption. , improve efficiency.

The above description is only an overview of the technical solutions of the present disclosure. In order to understand the technical means of the present disclosure more clearly, it can be implemented according to the contents of the description, and in order to make the above-mentioned and other purposes, features and advantages of the present disclosure more obvious and easy to understand , the following specific embodiments of the present disclosure are given.

Brief Description of Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the related technologies, the following briefly introduces the accompanying drawings used in the description of the embodiments or related technologies. Obviously, the drawings in the following description are the For the disclosed embodiments, for those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

FIG. 1 is a schematic diagram of a mode provided by an embodiment of the present disclosure;

2 is a flowchart of steps of a mode selection method provided by an embodiment of the present disclosure;

3 is a schematic diagram of another mode provided by an embodiment of the present disclosure;

4 is a flowchart of steps of another mode selection method provided by an embodiment of the present disclosure;

5 is a schematic diagram of a division provided by an embodiment of the present disclosure;

6 is a schematic diagram of a gradient direction of a computing block provided by an embodiment of the present disclosure;

FIG. 7 is a schematic diagram of a gradient direction of a PU provided by an embodiment of the present disclosure;

8 is a block diagram of a mode selection apparatus provided by an embodiment of the present disclosure;

FIG. 9 is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure; and

Figure 10 schematically shows a memory unit for holding or carrying program code implementing the method according to the present disclosure.

A detailed description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Furthermore, the embodiments of this disclosure and the features of the embodiments may be combined with each other without conflict.

First, the application scenarios involved in the embodiments of the present disclosure are described. Specifically, in the era of information explosion, digital video, as an important medium for people to obtain information in daily life, plays a very important role in the network society. However, the huge amount of information carried by the video information causes great difficulties in the transmission of video data. According to modern coding theories and methods, since the video data contains a lot of redundant information, the video data can be encoded and compressed to remove the redundant information to reduce the amount of data, thereby realizing the effective acquisition and utilization of video information.

In order to realize the encoding of video data, a variety of video encoding standards were born. For example, the H.264 video coding standard, the High Efficiency Video Coding (HEVC) standard, and the like. A video coding standard can provide multiple coding prediction modes. When using each mode to encode the same PU in a video image, the degree of image distortion and encoding bit rate will be different. Further, because different video images of different videos contain different image contents and are affected by the differences in contents, when different PUs are encoded using the same mode, the image distortion degree and encoding bit rate will also be different. Because, in order to balance the image distortion degree and the encoding bit rate during encoding, when encoding the video, it is necessary to select the mode with better encoding effect for the current PU, for example, select the optimal mode, and then use the mode Encode the current PU.

Taking the adoption of the HEVC standard as an example, the HEVC standard is a commonly used new-generation video coding standard, and its purpose is to address people's increasing demands for visual and auditory quality. The intra-frame prediction technology it provides is designed to eliminate spatial correlation in video data. According to the principle that the spatially closer pixels have a stronger correlation, it can use the principle of weighting adjacent pixels to predict the current pixel according to the correlation between the pixels. In this way, only the prediction residual needs to be If the difference is transformed and quantized, the amount of data to be transmitted can be greatly reduced. In order to achieve better coding performance, the HEVC standard provides a large number of modes. Specifically, the modes provided by the HEVC standard include two non-angle modes and 33 angle modes. By way of example, FIG. 1 is a schematic diagram of a mode provided by an embodiment of the present disclosure. As shown in FIG. 1 , “0Intra_Planar” and “1Intra_DC” respectively represent two non-angular modes: planar mode and DC mode, and 2 to 34 respectively represent 33 Medium angle mode. In an encoding scenario where the HEVC standard is used for video encoding, it is necessary to select a candidate mode from the 35 modes for each PU, and then select the optimal mode based on the candidate mode.

Further, in the related art, it is often necessary to traverse these 35 modes, calculate the rate-distortion cost of the current PU in each mode, and then select some modes as candidate modes according to the rate-distortion cost corresponding to each mode. n patterns as candidate patterns. Since the HEVC standard is used for video coding, a quad-tree structure is often used to divide the video image to divide a plurality of PUs. In this division method, a large number of PUs are often divided into a frame of video image. If the candidate modes are screened in this way, it will lead to the need to traverse 35 modes for a large number of PUs in the encoding process, resulting in a large amount of calculation and a long time-consuming, making the encoder load heavier and occupied for a long time. , which further limits the performance of the encoder.

To this end, an embodiment of the present disclosure proposes a mode selection method to simplify screening.

FIG. 2 is a flowchart of steps of a mode selection method provided by an embodiment of the present disclosure. As shown in FIG. 2 , the method may include:

Step 201: From the angle modes provided by the preset coding standard, select m angle representative modes; the m is less than the number of angle modes provided by the preset coding standard.

In this embodiment of the present disclosure, the preset encoding standard may be a standard for encoding the video to be encoded. Specifically, the preset coding standard may be selected according to actual requirements, for example, the preset coding standard may be the HEVC standard.

Further, since there are often a large number of angle modes in the modes provided by video coding, and in the angle modes provided by the preset coding standard, the angles represented by different angle modes are different, but the directions indicated by the angle modes with approximate angles are generally similar. For example, as can be seen from FIG. 1 , the angle pattern 11 indicating the horizontal direction is compared with other angle patterns adjacent to both sides thereof, such as the angle pattern 10 and the angle pattern 12 , the horizontal direction indicated by the angle pattern 11 is relatively The horizontal directions indicated by the angle patterns 10 and 12 are more precise, but all three generally point to the horizontal direction. For another example, the angle pattern 26 indicating the vertical direction is compared with other angle patterns adjacent to both sides thereof, such as the angle pattern 25 and the angle pattern 27, the vertical direction indicated by the angle pattern 26 is compared with that of the angle pattern 25. The vertical direction indicated by , 27 is more accurate, but the three generally point to the horizontal direction. Therefore, in this step, a part of the angle pattern can be selected as the angle representative pattern first, and an angle representative pattern represents that the angle can represent other patterns around the direction indicated by the pattern, that is, represents an angle pattern interval, and then realizes the use of these m The angle representative mode represents the angle mode provided by the preset coding standard.

Wherein, the specific value of m can be set according to the actual situation. For example, since the larger m is, the less representative modes of angles are selected, and the fewer modes that need to be processed subsequently, that is, the less computation is required, but correspondingly, the representativeness of the representative modes of angles will be weaker, and the screening The precision will be less. Conversely, the smaller m is, the more angle representative modes are selected, and the more modes that need to be processed subsequently, that is, the more calculation amount will be, but correspondingly, the representativeness of the angle representative mode will be stronger, and the screening accuracy will be stronger. will be higher. Therefore, in the embodiment of the present disclosure, a matching value may be selected as m according to the processing capability of the encoder and the requirement for the screening accuracy. Further, since m is less than the number of angle modes provided by the preset coding standard, it can be ensured to a certain extent that the mode of determining the candidate mode by selecting part of the angle modes as the angle representative mode results in a small amount of calculation and consumes less energy. The time is small, and the calculation efficiency is improved to a certain extent.

Step 202: Use the direction indicated by the angle representative mode as an alternative direction, and determine the first corresponding to the current PU from the alternative directions according to the coding effect parameters of the current prediction unit PU in each of the angle representative modes. direction; the angle corresponding to the first direction represents that the encoding effect parameter in the mode satisfies the first preset condition.

In this embodiment of the present disclosure, an angle representing the direction represented by the pattern may be used as a candidate direction, and then m candidate directions are obtained. As an example, it is assumed that 4 angle representation modes are selected: 10, 26, 2, and 18, and their respective directions may be: horizontal direction, vertical direction, lower-left direction, and upper-left direction. Then correspondingly, these four directions can be determined as candidate directions. It should be noted that the direction represented by each angle represented by the mode can be a relative direction. In modes with different representations, the direction corresponding to the same angle represented by the mode can be different, and the direction represented by the angle represented by the mode can also be extended in the opposite direction. direction. But the relative relationship of the directions represented by the various angles representing the mode is determined. For example, FIG. 3 is a schematic diagram of another mode provided by an embodiment of the present disclosure. As shown in FIG. 3 , the direction corresponding to the angle representative mode 18 at this time is the lower right direction, that is, the direction represented by 18 can be understood as the lower right direction direction.

Further, the encoding effect parameter of the current PU in the angle representative mode may be a parameter that can reflect the degree of image distortion and the size of the encoding bit rate for encoding the current PU using the angle representative mode. Because the main purpose of video coding is often to reduce the bit rate as much as possible under the condition of ensuring a certain video quality, or reduce the distortion as much as possible under the condition of ensuring a certain bit rate. Therefore, the rate-distortion cost can be used as a coding effect parameter, thereby reflecting the effect of coding the current PU by using the angle representative mode.

The first direction may be an alternative direction with better coding effect parameters in the corresponding angle representative mode. The first preset condition may be set according to the actual situation, as long as it is ensured that the encoding effect of the current PU is better by using the angle representative mode corresponding to the encoding effect parameter that satisfies the first preset condition. For example, the first preset condition may be that the encoding effect parameter is smaller than the preset parameter threshold, or the encoding effect represented by the encoding effect parameter is the best. Taking the coding effect parameter as the rate-distortion cost parameter as an example, since the smaller the rate-distortion cost, the better the coding performance, therefore, the candidate direction whose corresponding rate-distortion cost is less than the preset parameter threshold can be determined as the first direction, or is to determine the corresponding candidate direction with the smallest rate-distortion cost as the first direction, that is, the direction with the smallest rate-distortion cost in the candidate directions is used as the first direction.

Step 203: Select a candidate direction from the candidate directions according to the first direction; the candidate directions include at least the first direction.

In the embodiment of the present disclosure, since the encoding effect parameter in the angle representative mode corresponding to the first direction satisfies the first preset condition, that is, when the current PU is encoded using the angle representative mode corresponding to the first direction, the encoding effect is better. Therefore, it can be considered that a mode with better coding effect for the current PU is likely to exist in the modes around the first direction. Accordingly, the candidate direction may be selected from the candidate directions according to the first direction. Specifically, according to the first direction, a direction with a higher probability that a mode with better coding effect exists among the surrounding modes may be selected as a candidate direction.

Step 204: Based on the patterns around the candidate direction, select a candidate pattern from patterns provided by the preset coding standard.

In the embodiment of the present disclosure, since the candidate direction is determined according to the first direction and includes at least the first direction, it can be considered that there is a mode with better coding effect for the current PU in the modes around the candidate direction. Accordingly, the candidate mode may be selected from the modes provided by the preset coding standard based on the modes around the candidate direction. The angle patterns around the candidate direction may be partial angle patterns provided by the preset coding standard, that is, the number of angle patterns around the candidate direction is smaller than the number of angle patterns provided by the preset coding standard. In this way, the candidate modes included in the angle modes can be determined only by calculating part of the angle modes, thereby reducing the amount of calculation to a certain extent.

The mode selection method provided by the embodiment of the present disclosure selects m angle representative modes from the angle modes provided by the preset coding standard, where m is smaller than the number of angle modes provided by the preset coding standard, and takes the direction represented by the angle representative mode as an alternative direction, and according to the encoding effect parameters of the current prediction unit PU in each angle representative mode, select a first direction corresponding to the current PU from the alternative directions, and then select a candidate direction from the alternative directions according to the first direction, The candidate directions include at least the first direction. Finally, according to the candidate direction, a candidate mode is selected from the modes provided by the preset coding standard. In this way, the candidate direction in which the optimal mode may be located is determined by first selecting the angle representative mode, and the candidate mode is only selected in the mode around the candidate direction, that is, the direction with a small probability of having a mode with better coding effect is excluded, When screening for multiple angle modes, only the modes around the candidate directions, that is, some modes, are used as the basis for screening. In this way, the number of modes that need to be judged when screening candidate modes can be reduced to a certain extent, thereby reducing the amount of calculation and consumption. , improve efficiency.

FIG. 4 is a flowchart of steps of another mode selection method provided by an embodiment of the present disclosure. As shown in FIG. 4 , the method may include:

Step 301: From the angle modes provided by the preset coding standard, select m angle representative modes; the m is less than the number of angle modes provided by the preset coding standard.

Specifically, this step can be implemented by the following operations: starting from the preset angle mode, selecting m angle modes at equal intervals from the angle modes provided by the preset coding standard in sequence according to the second preset step size; The m angle patterns of are taken as the angle representative patterns.

When making a selection, the angle pattern at the starting point can be extracted, and then starting from the first extracted angle pattern, the angle pattern reached after the second preset step is extracted, and then starting from the currently extracted angle pattern , extracting the angle pattern reached after the second preset step, and so on, until enough m angle pattern positions are extracted. The preset angle mode and the second preset step size may be preset according to actual conditions. For example, the angle mode 2 may be used as the preset angle mode, or the angle mode 34 may be used as the preset angle mode, and the second preset step size is set to 8. Take the preset angle mode as angle mode 2, the second preset step size is 8, and m is 4 as an example, you can first extract angle mode 2, and then start from angle mode 2, extract the position reached after 8 steps. The angle pattern 10, then, starting from the angle pattern 10, extract the angle pattern 18 where it reaches after 8 steps, and starting from the angle pattern 18, extract the angle pattern 26 where it reaches after 8 steps, and obtain four Angle mode: angle mode 2, angle mode 10, angle mode 18, angle mode 26. Finally, the four angle patterns can be determined as angle representative patterns. In the embodiment of the present disclosure, extraction is performed at equal intervals by the second preset step size, so that it can be ensured that the finally obtained angle representative patterns are evenly distributed in the angle patterns, thereby ensuring that each angle representative pattern can evenly represent its surroundings. angle mode, improving the reliability of selection based on the angle representative mode.

Step 302: Use the direction represented by the angle representative mode as a candidate direction, and calculate the rate-distortion cost of the current PU in each of the angle representative modes.

In this step, for each angle representative mode, the following operations may be performed respectively to calculate the rate-distortion cost of the current PU in each angle representative mode:

Use the coding parameters defined by the angle to represent the mode to calculate the absolute sum of coefficients (Sum of Absolute Transformed Difference, SATD) and the corresponding coding bit rate (R) after the Hardman transform of the residual of the current PU, and use SATD to represent the current The degree of distortion D of the PU at this angle represents the mode. Then bring the two into the following formula to get the rate-distortion cost (J(mode)) of the current PU in the representative mode of this angle:

J(mode)=SATD+λ*R;

Among them, λ is a preset coefficient.

Since the amount of computation for encoding using the mode is relatively large, compared to directly using the angle representative mode to encode the current PU, the distortion degree D of the current PU in the angle representative mode is obtained, and the current PU is calculated in combination with the distortion degree D. The way the PU represents the rate-distortion penalty in the mode at this angle. In this step, D is represented by SATD, so that the rate-distortion cost of the current PU in this angle representative mode can be calculated without encoding. Therefore, the calculation cost of the rate-distortion cost can be reduced to a certain extent, and the rate-distortion cost of the rate-distortion cost can be improved to a certain extent. Computational efficiency. Of course, the degree of distortion D can also be used directly for calculation, or other parameters can be used to indicate that D participates in the calculation, for example, using the sum of absolute error (Sum of Absolute Difference, SAD), the sum of squared differences (Sum of Squared Difference, SSD) , etc., which are not limited in the embodiments of the present disclosure.

Step 303: Determine the candidate direction corresponding to the angle representative mode with the smallest rate-distortion cost as the first direction.

In this step, the magnitude of the rate-distortion cost corresponding to each angle representative mode can be compared, and then the rate-distortion cost with the smallest value is selected. Then, the angle corresponding to the minimum rate-distortion cost is an alternative corresponding to the mode representative mode. The direction is determined as the first direction. For example, assuming that the candidate direction corresponding to the angle representative mode corresponding to the minimum rate-distortion cost is the horizontal direction, the horizontal direction may be determined as the first direction.

In the embodiment of the present disclosure, by selecting the candidate direction corresponding to the angle representative mode with the smallest rate-distortion cost as the first direction, that is, selecting the direction with the smallest rate-distortion cost, the selected first direction can be the one with the most encoding effect on the current PU. A good direction can further ensure the reliability of the candidate direction selected based on the first direction in the subsequent steps.

Step 304: According to the gradient strength of the current PU in each of the candidate directions, select a second direction corresponding to the current PU from the candidate directions; the gradient strength in the second direction satisfies the second preset condition.

In this step, the gradient may be a vector with magnitude and direction, and the direction of the gradient may be the direction indicating the change of the pixel value, that is, the direction of the texture of the image may be represented. The size can be the size of the gradient strength, which can characterize how fast the pixel value of the pixel changes in this direction. The gradient strength in the alternative direction can represent the magnitude of the gradient, that is, it can be used to represent how fast the pixel value of the pixel point in the current PU changes in the alternative direction. The smaller the gradient strength is, the smaller the speed is, that is, the slower the pixel value of the pixel point changes in the alternative direction. Wherein, the degree of change of the pixel value in the alternative direction may reflect the texture of the image texture in the current PU in the alternative direction. The slower the change, the stronger the texture. The stronger the texture is, the more regular the image texture in the current PU is in this direction, on the contrary, the weaker the texture is, the more chaotic the image texture in the current PU is in this direction.

Further, the second direction may be a corresponding candidate direction with a smaller gradient intensity, that is, a candidate direction with relatively slow change and strong texture of the image texture. The second preset condition may be set according to the actual situation, as long as it is ensured that the change of the pixel value in the direction satisfying the second preset condition is relatively slow. For example, the second preset condition may be that the gradient strength is smaller than the preset gradient threshold, or the gradient strength is the smallest, that is, the smallest gradient direction among the alternative directions is used as the second direction.

Specifically, this step can be implemented through the following steps 3041 to 3044:

Step 3041: Divide the current PU into at least two calculation blocks, wherein one of the calculation blocks includes at least two pixels.

In this step, the number of pixels included in the calculation block may be set according to the candidate direction. For example, in the case of one candidate direction, the set number is not less than 2, and if the candidate direction is two In the case that there are more than 3 alternative directions, the set number is not less than 3. In the case of more than 3 alternative directions, the set number is not less than 4. The shape of the calculation block may be a square, a rectangle, or the like. Not limited. Specifically, when dividing, starting from a preset position, for example, starting from the upper left corner of the current PU, at least two pixels are sequentially selected as a calculation block, and then multiple calculation blocks are obtained. For example, taking the current PU as an 8*8 area as an example, FIG. 5 is a schematic diagram of a division provided by an embodiment of the present disclosure. As shown in FIG. 5 , the forward direction area composed of every 16 pixels can be divided into one Calculation block, and then 4 calculation blocks are obtained, wherein the 4 calculation blocks are: 4*4 pixel blocks covered by slashes, 4*4 pixel blocks covered by black dots, 4*4 pixel blocks covered by horizontal lines and A 4*4 pixel block covered by vertical lines. It should be noted that, during division, it can be controlled that there is no overlap between the calculation blocks obtained by division. In this way, compared with the method of using the quad-tree structure for division, resulting in the occurrence of overlapping parts, the embodiment of the present disclosure can avoid the problem of repeated calculation due to the existence of repeated parts.

Step 3042: For any of the candidate directions, calculate the gradient strength of the calculation block in the candidate direction according to the pixel values of the pixel points included in each of the calculation blocks.

Specifically, calculating the gradient strength of the calculation block in the candidate direction is to use the candidate direction as the gradient direction, and calculate the gradient strength of the calculation block in the gradient direction. For any calculation block, the gradient strength of the calculation block in the alternative direction can be calculated by the following operations: according to the alternative direction, n pairs of pixel points are selected at intervals from the calculation block as reference pixel points pair; perform differential calculation based on the pixel points included in the reference pixel point pair; determine the result of the differential calculation as the gradient strength of the calculation block in the candidate direction.

In this step, n may be set according to the actual situation. The larger n is, the more reference pixel pairs are selected, and accordingly, the more accurate the gradient intensity of the calculation block calculated based on these reference pixel pairs is. Specifically, when selecting a reference pixel pair, two non-adjacent pixels may be selected at intervals in the alternative direction to form a reference pixel pair. Among them, the non-adjacent pixels may refer to the non-adjacent up, down, left and right, that is, to ensure that there are no pixels in the reference pixel pair that are 4 adjacent pixels to each other. The upper right and lower right are not adjacent, that is, it is ensured that there are no pixels that are 8 neighboring pixels to each other in the reference pixel pair. Due to the similarity between adjacent pixels, compared with the method of directly using adjacent pixels for calculation, in the embodiment of the present disclosure, by using pixels selected at intervals to form reference pixel pairs, it is possible to reduce the number of pixels to a certain extent. The interference caused by this similarity enables the gradient strength to better reflect the texture trend in the alternative direction, and to describe the texture more accurately, which in turn can improve the effect of the calculated gradient strength on the texture of the calculation block. Indicates precision.

Next, a difference calculation is performed based on the reference pixel pair. Among them, difference is also known as difference function or difference operation, and the result of difference can reflect a change between discrete quantities. Specifically, during the calculation, the absolute value of the pixel value difference between the two pixel points in each reference pixel point pair can be calculated, and then the sum of all absolute values can be calculated to obtain the difference result. Finally, the difference result in the alternative direction is used as the gradient strength of the calculation block in the alternative direction.

For example, it is assumed that the calculation block is a 4*4 calculation block divided in FIG. 5 , and n is 4. From any vertex of the calculation block as the origin, the pixels are numbered according to 0, 1, 2, and 3 along the horizontal X-axis direction and the vertical Y-axis direction respectively, then, the gradient strength of a calculation block in the horizontal direction GH can be expressed as:

G _H =|p(0,3)-p(0,1)|+|p(1,2)-p(1,0)|+|p(2,3)-p(2,1)| +|p(3,2)-p(3,0)|

The gradient strength G _v in the vertical direction can be expressed as:

G _V =|p(3,0)-p(1,0)|+|p(2,1)-p(0,1)|+|p(3,2)-p(1,2)| +|p(2,3)-p(0,3)|

The gradient strength G _rd in the lower right direction can be expressed as:

G _rd =|p(0,0)-p(2,2)|+|p(1,1)-p(3,3)|+|p(2,0)-p(3,1)| +|p(0,2)-p(1,3)|

The gradient strength G _ld in the lower left direction can be expressed as:

G _ld =|p(0,3)-p(2,1)|+|p(1,2)-p(3,0)|+|p(0,1)-p(1,0)| +|p(2,3)-p(3,2)|

Further, FIG. 6 is a schematic diagram of a gradient direction of a calculation block provided by an embodiment of the present disclosure. It can be seen that the gradient direction of the calculation block includes the horizontal direction shown in (3a), the vertical direction shown in (3b), The lower right direction shown in 3c) and the lower left direction shown in (3d).

Step 3043: Calculate the gradient strength corresponding to the current PU in the candidate direction according to the gradient strength of the calculation block in the candidate direction.

Specifically, the sum of the gradient strengths of the calculation blocks of all calculation blocks in the candidate direction may be calculated, and the sum of the gradient strengths may be determined as the gradient strength corresponding to the current PU in the candidate direction. Further, FIG. 7 is a schematic diagram of a gradient direction of a PU provided by an embodiment of the present disclosure, and it can be seen that the gradient directions of the PU include the horizontal direction shown in (4a), the vertical direction shown in (4b), and (4c) The lower right direction shown and the lower left direction shown by (4d).

In the embodiment of the present disclosure, by dividing the current PU into multiple calculation blocks, and taking the calculation blocks as the gradient calculation objects, it is only necessary to calculate the gradient strengths of these calculation blocks in the alternative directions, and the calculation can be done based on the gradient strengths of these calculation blocks. The gradient strength corresponding to the current PU in the alternative direction is obtained. In this way, the gradient calculation process can be simplified to a certain extent, the calculation amount caused by the gradient calculation can be reduced, and the calculation efficiency can be improved. As an example, taking the current PU in FIG. 5 as an example, if a conventional calculation method is used and no division is performed, it is necessary to calculate the gradient strengths of the 64 pixels in the alternative directions for the 64 pixels respectively, and obtain 64 gradient strength. Then the 64 gradient strengths are calculated to obtain the gradient strength of the current PU in the alternative direction. For example, the sum of the gradient strengths of all pixels in the alternative direction is calculated, that is, the calculation SAG=Σ(G(x, y)). However, by using the calculation method provided by the embodiment of the present disclosure, it is only necessary to calculate the gradient intensities of the four calculation blocks in the alternative directions, and obtain four gradient intensities. Then, these four gradient strengths are calculated to obtain the gradient strength of the current PU in the alternative direction. It can be seen that, compared with the prior art, in which the pixel points in the current PU are differentiated point by point in the alternative directions, the method provided by the embodiment of the present disclosure can simplify the gradient calculation process and reduce the burden caused by the gradient calculation. the amount of computation, thereby improving the computational efficiency.

Meanwhile, in the embodiment of the present disclosure, when calculating the gradient strength of the calculation block, the pixel points are extracted at intervals to participate in the calculation, which can reduce the number of pixel points participating in the calculation, thereby further reducing the amount of calculation. At the same time, compared with the method of performing the difference point by point in the prior art, in the embodiment of the present disclosure, the method of extracting pixel points at intervals to participate in the calculation can also avoid that the difference between the adjacent points is too small, resulting in the final calculation result being unable to be calculated. Accurately reflect the texture of the problem. Of course, the gradient intensity may also be calculated in other ways, for example, by calculating the derivative by using a two-dimensional discrete function, and so on, which is not limited in this embodiment of the present disclosure.

It should be noted that, in this embodiment of the present disclosure, before calculating the gradient strength of the current PU, that is, before performing texture detection, low-pass filtering may also be performed on the video image to reduce interference information contained in the video image, thereby improving texture detection effect.

Step 3044: Determine the corresponding candidate direction with the smallest gradient strength as the second direction.

In this step, the magnitudes of the gradient strengths corresponding to each candidate direction may be compared, and then the gradient strength with the smallest value is selected, and then the candidate direction corresponding to the minimum gradient strength is determined as the second direction. Exemplarily, assuming that the candidate direction corresponding to the smallest gradient strength is the horizontal direction, the horizontal direction may be determined as the second direction. In the embodiment of the present disclosure, by selecting the candidate direction with the smallest gradient as the second direction, that is, selecting the minimum gradient direction, the selected second direction can be the direction with the strongest texture of the current PU, and then the second direction can be ensured reliability.

Step 305: Select the candidate direction from the candidate directions according to the relative relationship between the first direction and the second direction.

In practical application scenarios, the texture direction of the video image, that is, the direction with strong texture, is often similar to the direction of the mode with better coding effect, for example, the direction of the optimal mode is similar, and the direction of the mode with better coding effect is similar. Often also has a strong texture. And when using different modes to encode an image, the encoding effect will be affected by the image texture, and the second direction can represent the direction with strong regularity of the image texture of the current PU. Therefore, in this embodiment of the present disclosure, the first direction can be combined. and the second direction, that is, the candidate direction is selected by combining the minimum cost direction and the minimum gradient direction.

Specifically, if the first direction is parallel to the second direction, the candidate direction indicated by the first direction is determined as the candidate direction. If the first direction is parallel to the second direction, it means that the minimum cost direction and the minimum gradient direction are consistent, the image represented by the current PU has strong texture, the texture detection result is reliable, and the reference is high, that is, the determined first Both directions are reliable. However, since the first direction is similar to the second direction, if the second direction is reliable, the first direction can be considered to be reliable, and then only the candidate direction represented by the first direction can be determined as a candidate direction. That is, only the second direction is determined as a candidate direction. In the embodiment of the present disclosure, when the minimum cost direction and the minimum gradient direction are consistent and have a high degree of reliability, the candidate direction represented by the two is selected as the candidate direction, which can ensure that the candidate direction can be more accurate in the future based on the candidate direction. While screening candidate modes, reduce the number of candidate directions as much as possible, thereby reducing the amount of subsequent processing.

Further, if the two are not parallel, it can be further determined whether the two are perpendicular, and if the first direction is perpendicular to the second direction, at least two of the candidate directions are selected as the candidate directions. If the first direction is perpendicular to the second direction, it means that there is a big difference between the minimum cost direction and the minimum gradient direction, the texture of the image represented by the current PU is weak, and the result of texture detection is less reliable, that is, it can be considered that The reliability of the first direction is low, and the reference is low and high. And the weak texture is often caused by the complex texture of the image, or the lack of texture details. In this case, if the candidate mode is only determined in a single direction, it may lead to the final selection based on the candidate mode. Better to encode the current PU. Therefore, multiple candidate directions can be selected as candidate directions, so that it can be ensured that the optimal mode determined based on the candidate modes screened in the candidate direction can better encode the current PU, thereby ensuring the subsequent encoding effect. Meanwhile, the selected at least two candidate directions may include the candidate direction represented by the first direction. Since the first direction is an alternative direction with better coding effect parameters in the corresponding angle representative mode, selecting the first direction can further ensure the optimal selection determined by the candidate mode selected based on the candidate direction to a certain extent. The encoding effect of the mode. Further, all candidate directions may be selected as candidate directions. Taking 4 candidate directions as an example, the 4 candidate directions can be determined as candidate directions, and candidates can be made in these 4 directions. In this way, by determining all the candidate defense lines as candidate directions, the coverage direction can be greatly improved, thereby ensuring the coding effect of the optimal mode determined based on the candidate modes screened in the candidate direction subsequently.

Further, if an included angle of a preset angle is formed between the first direction and the second direction, the candidate direction represented by the first direction and the candidate direction represented by the second direction are determined. is a candidate direction; wherein, the preset angle is an angle other than 0 degrees and 90 degrees.

If a predetermined angle is formed between the first direction and the second direction, that is, the two directions are neither parallel nor perpendicular. It means that there are some differences between the minimum cost direction and the minimum gradient direction, but the result of texture detection has a certain degree of reliability and has a certain reference. Therefore, the candidate directions represented by the first direction and the candidate directions represented by the second direction may be selected as candidate directions. In this way, by simultaneously selecting the candidate directions represented by the first direction and the second direction as the candidate directions, the problem of low reference can be compensated to a certain extent, thereby improving the accuracy of the candidate directions.

In the embodiment of the present disclosure, by combining the first direction and the second direction, that is, combining the minimum cost direction and the minimum gradient direction, a double verification judgment is performed, the minimum gradient direction is used to assist the judgment, and according to the judgment result, the candidate direction is selected in a targeted manner, Further, the accuracy of the candidate directions that are initially screened can be improved. Since the candidate directions of the preliminary screening have a great influence on the mode finally selected by the scheme, in the embodiment of the invention, by improving the accuracy, the overall accuracy of the scheme can be ensured.

Step 306: For any of the candidate directions, centering on the angle pattern corresponding to the candidate direction, according to the first preset step size, select p patterns at equal intervals from the patterns on both sides of the corresponding angle pattern. , as the first alternative mode.

In this step, the first preset step size may be preset according to the actual situation, and the first preset step size may be greater than 1. By setting the first preset step size greater than 1, it can be ensured that there are more candidate directions. In this case, the number of candidate modes finally selected will not be too many, which can avoid a large amount of calculation. For example, the first preset step size may be 2. Further, the specific value of p may also be preset according to the actual situation, for example, p may be 4. During specific selection, the angle pattern corresponding to the candidate direction may be taken as the center, and the corresponding angle pattern and the patterns on both sides thereof may be selected as objects. Start the selection at equal intervals with the first preset step.

Assume that the first preset step size is 2 and p is 4. When the candidate direction is the horizontal direction, and the angle pattern corresponding to the candidate direction is the angle pattern 10, when selecting p patterns at equal intervals from the patterns on both sides of the corresponding angle pattern, the first angle pattern 7 or the angle pattern 13 can be used as the starting point. , first extract the angle pattern at the starting point, then start from the extracted angle pattern, extract the angle pattern reached after the first preset step, and so on, the first alternative pattern can be obtained: angle pattern 7 , Angle Mode 9, Angle Mode 11, Angle Mode 13. Correspondingly, when the candidate direction is the vertical direction and the angle mode corresponding to the candidate direction is the angle mode 26, p modes are selected at equal intervals from both sides of the corresponding angle mode, and the first candidate mode can be obtained: the angle mode 23, Angle Mode 25, Angle Mode 27, Angle Mode 29. When the candidate direction is the lower right direction and the angle mode corresponding to the candidate direction is the angle mode 18, select p modes at equal intervals from both sides of the corresponding angle mode, and the first candidate mode can be obtained: angle mode 15, angle mode 17. Angle Mode 19, Angle Mode 21. When the candidate direction is the lower left direction and the angle mode corresponding to the candidate direction is angle mode 2, select p modes at equal intervals from both sides of the corresponding angle mode, and the first candidate modes can be obtained: angle mode 3, angle mode 5 , Angle Mode 31, Angle Mode 33. In the embodiment of the present disclosure, selection and extraction are performed at equal intervals by the first preset step size, so that the final obtained first candidate mode can be uniformly distributed, thereby facilitating further selection on the basis of the first candidate mode in the future.

Step 307: Determine a first candidate mode according to the first candidate mode, and determine a non-angular mode provided by the preset coding standard as a second candidate mode.

Specifically, the operation of determining the first candidate mode according to the first candidate mode may be implemented by the following operations: calculating the rate-distortion cost of the current PU in each of the first candidate modes; The first candidate mode corresponding to the distortion cost is used as the second candidate mode; the angle modes adjacent to both sides of the second candidate mode are determined as the first candidate mode.

For the specific implementation manner of calculating the rate-distortion cost of the current PU in each of the first alternative modes, reference may be made to the relevant description of calculating the rate-distortion cost in the preceding steps, which is not repeated in this embodiment of the present disclosure. Further, the specific value of q is set according to the actual situation. For example, q may be 2. The rate-distortion costs may be sorted in descending order, and then the first candidate mode corresponding to the first two rate-distortion costs may be selected as the second candidate mode. Assuming that the second candidate angle modes are angle mode 7 and angle mode 13 , the selected first candidate modes may be angle mode 6 , angle mode 8 , angle mode 12 , and angle mode 14 . It should be noted that, after the first candidate mode is determined, deduplication processing may be performed on the first candidate mode, so as to reduce the number of the first candidate mode, thereby reducing the subsequent calculation amount. Alternatively, the first candidate mode can also be directly determined as the first candidate mode, that is, the operation of further screening is omitted, thereby reducing the amount of calculation.

Further, since the directionality of the non-angle mode is very weak, it is difficult to be covered in the selection process of the mode around the candidate direction. Therefore, in this step, the non-angle mode can be directly determined as the second candidate mode, that is, It is ensured that non-angle mode can also be added to screening, which can improve the screening efficiency while ensuring comprehensive coverage and comprehensive screening, thereby improving the accuracy of screening results. By way of example, taking the preset coding standard as the HEVC standard as an example, we may determine the planar mode and the dc mode as the second candidate mode.

Step 308: Determine the first candidate mode and the second candidate mode as the candidate mode, and add the angle representative mode and the first candidate mode to the candidate mode.

In the embodiment of the present disclosure, the first candidate mode and the second candidate mode are obtained through multi-level screening, and finally the first candidate mode and the second candidate mode are determined as candidate modes, which can ensure the accuracy of the final selected candidate mode to a certain extent. sex. Further, by adding the angle representative mode and the first candidate mode to the candidate modes, that is, the angle representative mode is used as the first-level candidate mode, and the first candidate mode and the non-angle mode are used as the second-level candidate mode. In the candidate mode, the first candidate mode is regarded as the third-level candidate mode. By adopting the multi-level candidate mode, layer-by-layer screening can be avoided, resulting in the final screening result falling into a local optimum, which in turn leads to the problem of unsatisfactory screening results.

Further, it should be noted that, after the candidate mode is determined in this embodiment of the present disclosure, the following operations may be performed: determine the rate-distortion cost corresponding to the candidate mode; if the rate-distortion cost corresponding to the candidate mode is greater than a preset value If the rate-distortion cost threshold is set, the candidate mode is eliminated. Wherein, if the corresponding rate-distortion cost has been calculated for the candidate mode in the foregoing steps, the corresponding rate-distortion cost can be obtained directly from the calculation result of the foregoing steps, thus reducing the amount of computation. If not, the calculation can be performed with reference to the description of the calculation method in the preceding steps. Further, the preset rate-distortion cost threshold may be set according to the rate-distortion cost corresponding to a mode that does not become the optimal mode in an actual application scenario. Alternatively, it can also be set according to the minimum rate-distortion cost calculated during the experiment, for example, the threshold is set to 1.2 times the minimum rate-distortion cost. Wherein, the minimum rate-distortion cost may be based on the rate-distortion cost calculated by SATD.

Correspondingly, if the rate-distortion cost corresponding to the candidate mode is greater than the preset rate-distortion cost threshold, it can be considered that the candidate mode has a high probability that it will not become the optimal mode, so it can be eliminated. In this way, by selectively eliminating the candidate modes, the number of candidate modes can be reduced, thereby reducing the amount of calculation in subsequent screening from the candidate modes.

In the actual application scenario, after the above operation, that is, after the rough mode selection (Rough Mode Decision, RMD) stage is completed, the stage of adding the most probable modes (Most Possible Modes, MPM) can be continued, and the fine mode selection (Fine Mode Decision, FMD) stage can be continued. )stage. Specifically, since there is often a strong correlation between adjacent PUs, the optimal mode of the coded adjacent PUs is also valuable for the current block. Therefore, in this embodiment of the present disclosure, a candidate list may be generated based on the previously selected candidate mode, the MPM mode may be determined according to the optimal mode of the coded PU around the current PU, and then the MPM mode may also be added to the candidate list for subsequent selection. The MPM mode may be determined from the optimal modes of the coded adjacent PUs, eg, the upper and left coded adjacent PUs, according to the spatial correlation of the coded adjacent PUs. The specific number of specific MPM modes can be set according to the actual situation. For example, in the simulation platform HM16.0, the number of MPMs can be 3. When determining the specific number, the coded adjacent PUs on the upper and left sides can be obtained first. Then, according to the characteristics of the two optimal modes and the relationship between them, the three MPMs are assigned values in turn, and then three MPM modes are obtained.

Next, the rate-distortion cost based on RDO can be calculated for all modes in the candidate list in turn, and the prediction mode with the smallest rate-distortion cost is selected as the optimal mode. Among them, the solution of the RDO cost includes a complete encoding process, so the optimal mode can be accurately selected, but this process requires DCT transformation of the residual of the predicted pixels, quantization to obtain residual coefficients, and entropy encoding to obtain the prediction mode. On the other hand, it is necessary to inverse quantize and inverse transform the residual coefficient to obtain the reconstructed image to obtain the distortion degree. Finally, the rate-distortion cost of the prediction mode can be calculated according to the number of bits and the distortion degree, and the complexity is extremely high. Therefore, in practical applications, other methods can also be used to calculate the rate-distortion cost and select it, thereby optimizing the complexity of mode selection. It should be noted that, before the FMD stage, the pattern included in the candidate list may also be deduplicated, so as to reduce the number of objects to be calculated in the FMD stage, thereby reducing the amount of calculation.

The mode selection method provided by the embodiment of the present disclosure selects m angle representative modes from the angle modes provided by the preset coding standard, where m is smaller than the number of angle modes provided by the preset coding standard, and takes the direction represented by the angle representative mode as Alternative directions, and according to the coding effect parameters of the current prediction unit PU in each angle representative mode, select the first direction corresponding to the current PU from the alternative directions, and according to the gradient strength of the current PU in each alternative direction, from the alternative directions In the selection direction, select the second direction corresponding to the current PU. Then, combining these two directions, a candidate direction is selected from the candidate directions. Finally, according to the candidate directions, the first candidate mode is determined from the angle modes, the non-angle mode is determined as the second candidate mode, and finally the first candidate mode and the second candidate mode are determined as the candidate modes. In this way, by first selecting the angle representative mode to determine the possible candidate direction of the optimal mode, and selecting the candidate mode only on the modes around the candidate direction, the number of modes that need to be judged when screening the candidate modes can be reduced to a certain extent, and then Reduce the amount of calculation and time-consuming, and improve efficiency. At the same time, due to the weak directivity of the non-angle mode, it is difficult to be covered in the selection process for the surrounding modes of the candidate direction. Therefore, by directly determining the non-angle mode as the second candidate mode, that is, let the non-angle mode Screening can also be added to ensure comprehensive coverage while improving screening efficiency, thereby improving the accuracy of screening results. Further, by combining the first direction and the second direction, that is, combining the minimum cost direction and the minimum gradient direction, a double verification judgment is performed, the minimum gradient direction is used to assist the judgment, and according to the judgment result, the candidate direction can be selected in a targeted manner, which can improve the The accuracy of the initially screened candidate directions.

FIG. 8 is a block diagram of a mode selection apparatus provided by an embodiment of the present disclosure. As shown in FIG. 8 , the apparatus 40 may include:

The first selection mode 401 is used to select m angle representative modes from the angle modes provided by the preset coding standard; the m is less than the number of angle modes provided by the preset coding standard;

The first determination mode 402 is configured to use the direction represented by the angle representative mode as an alternative direction, and determine the current direction from the alternative directions according to the coding effect parameters of the current prediction unit PU in each of the angle representative modes. The first direction corresponding to the PU; the angle corresponding to the first direction represents the encoding effect parameter in the mode meeting the first preset condition;

A second selection mode 403, configured to select a candidate direction from the candidate directions according to the first direction; the candidate directions include at least the first direction;

A third selection mode 404 is configured to select a candidate mode from modes provided by the preset coding standard based on modes around the candidate direction.

Optionally, the second selection mode 403 is specifically used for:

According to the gradient strength of the current PU in each of the candidate directions, a second direction corresponding to the current PU is selected from the candidate directions; the gradient strength in the second direction satisfies a second preset condition.

The candidate direction is selected from the candidate directions according to the relative relationship between the first direction and the second direction.

Optionally, the second selection mode 403 is also specifically used for:

The current PU is divided into at least two calculation blocks; wherein one of the calculation blocks includes at least two pixel points.

For any of the alternative directions, the gradient strength of the calculation block in the alternative direction is calculated according to the pixel values of the pixel points included in each of the calculation blocks.

According to the gradient strength of the calculation block in the candidate direction, the gradient strength corresponding to the current PU in the candidate direction is calculated.

The corresponding candidate direction with the smallest gradient intensity is determined as the second direction.

Optionally, the second selection mode 403 is also specifically used for:

For any of the calculation blocks, according to the alternative direction, n pairs of pixel point pairs are selected from the calculation block at intervals as reference pixel point pairs.

The difference calculation is performed based on the pixels included in the reference pixel pair.

The result of the difference calculation is determined as the gradient strength of the calculation block in the candidate direction.

Optionally, the second selection mode 403 is also specifically used for:

If the first direction is parallel to the second direction, the candidate direction indicated by the first direction is determined as the candidate direction.

If the first direction is perpendicular to the second direction, at least two of the candidate directions are selected as the candidate directions.

If an included angle of a preset angle is formed between the first direction and the second direction, the candidate direction represented by the first direction and the candidate direction represented by the second direction are determined as the candidate direction; wherein, the preset angle is an angle other than 0 degrees and 90 degrees.

Optionally, the second selection mode 403 is also specifically used for:

All the candidate directions are selected and determined as the candidate directions.

Optionally, the third selection mode 404 is specifically used for:

For any of the candidate directions, with the angle pattern corresponding to the candidate direction as the center, according to the first preset step size, select p patterns at equal intervals from the patterns on both sides of the corresponding angle pattern, as the first An alternative mode.

A first candidate mode is determined according to the first candidate mode, and a non-angular mode provided by the preset coding standard is determined as a second candidate mode.

The first candidate mode and the second candidate mode are determined as the candidate modes.

Optionally, the third selection mode 404 is also specifically used for:

Calculate the rate-distortion cost of the current PU in each of the first candidate modes.

The first candidate mode corresponding to the first q largest rate-distortion costs is used as the second candidate mode.

An angle mode adjacent to both sides of the second candidate mode is determined as the first candidate mode.

Optionally, the device 40 further includes:

An adding module is used for adding the angle representative mode and the first candidate mode to the candidate mode.

Optionally, the device 40 further includes:

The second determination module is configured to determine the rate-distortion cost corresponding to the candidate mode.

A rejection module, configured to reject the candidate mode if the rate-distortion cost corresponding to the candidate mode is greater than a preset rate-distortion cost threshold.

Optionally, the first selection mode 401 is specifically used for:

Taking the preset angle mode as a starting point, m angle modes are selected from the angle modes provided by the preset coding standard at equal intervals in sequence according to the second preset step size.

The selected m angle patterns are used as the angle representative patterns.

Optionally, the first determination mode 402 is specifically used for:

A rate-distortion cost for the current PU in each of the angle representation modes is calculated.

The angle with the smallest rate-distortion cost represents the candidate direction corresponding to the mode, and is determined as the first direction.

To sum up, the mode selection device provided by the embodiment of the present disclosure selects m angle representative modes from the angle modes provided by the preset coding standard, where m is less than the number of angle modes provided by the preset coding standard, and the angle representative mode is selected. The direction represented by the mode is used as an alternative direction, and according to the coding effect parameters of the current prediction unit PU in each angle representative mode, the first direction corresponding to the current PU is selected from the alternative directions, and then according to the first direction, from the alternative direction Select candidate directions from among the candidate directions, and the candidate directions include at least the first direction. Finally, according to the candidate direction, a candidate mode is selected from the modes provided by the preset coding standard. In this way, the candidate direction in which the optimal mode may be located is determined by first selecting the angle representative mode, and the candidate mode is only selected in the mode around the candidate direction, that is, the direction with a small probability of having a mode with better coding effect is excluded, When screening for multiple angle modes, only the mode around the candidate direction, that is, the partial mode, is the basis for screening. In this way, the number of modes that need to be judged when screening the candidate mode can be reduced to a certain extent, thereby reducing the amount of calculation and consumption. , improve efficiency.

For the above-mentioned apparatus embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and reference may be made to the partial descriptions of the method embodiments for related parts.

The device embodiments described above are only illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in One place, or it can be distributed over multiple network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

Optionally, an embodiment of the present disclosure further provides an electronic device, the electronic device may include: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the When the processor executes the program, each process of the above mode selection method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, details are not repeated here.

Optionally, an embodiment of the present disclosure further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, each process of the foregoing mode selection method embodiment is implemented, and can To achieve the same technical effect, in order to avoid repetition, details are not repeated here.

For example, FIG. 9 shows a schematic structural diagram of an electronic device. The electronic device includes a processor 510 and a memory 520 . The memory 520 may be something such as flash memory, electrically erasable programmable read-only memory, hard disk, random access memory (RAM), magnetic disk, optical disk, or read-only memory (ROM) or the like electronic storage. Memory 520 has storage space 530 for program codes. These program codes can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks, portable or fixed storage units as described with reference to FIG. 10 . The storage unit may have storage segments, storage spaces, etc. arranged similarly to the memory 520 in the electronic device of FIG. 9 . Typically, the storage unit includes computer readable code 1031' which, when executed by a computing processing device, causes the computing processing device to perform the various steps in the methods described above.

Claims (16)

1. A mode selection method, characterized in that the method comprises:

   From the angle modes provided by the preset coding standard, select m angle representative modes; the m is less than the number of angle modes provided by the preset coding standard;

   Taking the direction represented by the angle representative mode as an alternative direction, and determining the first direction corresponding to the current prediction unit from the alternative directions according to the coding effect parameters of the current prediction unit in each of the angle representative modes ; The encoding effect parameter under the angle representative mode corresponding to the first direction satisfies the first preset condition;

   A candidate direction is selected from the candidate directions according to the first direction; the candidate directions include at least the first direction;

   Based on the patterns around the candidate direction, a candidate pattern is selected from patterns provided by the preset coding standard.
2. The method according to claim 1, wherein the selecting a candidate direction from the candidate directions according to the first direction comprises:

   According to the gradient strength of the current prediction unit in each of the candidate directions, a second direction corresponding to the current prediction unit is selected from the candidate directions; the gradient strength in the second direction satisfies a second preset condition ;

   The candidate direction is selected from the candidate directions according to the relative relationship between the first direction and the second direction.
3. The method according to claim 2, wherein the second direction corresponding to the current prediction unit is selected from the candidate directions according to the gradient strength of the current prediction unit in each of the candidate directions, include:

   dividing the current prediction unit into at least two calculation blocks; wherein one of the calculation blocks includes at least two pixels;

   For any of the alternative directions, according to the pixel values of the pixel points included in each of the calculation blocks, calculate the gradient strength of the calculation block in the alternative direction;

   According to the gradient strength of the calculation block in the candidate direction, calculate the gradient strength corresponding to the current prediction unit in the candidate direction;

   The corresponding candidate direction with the smallest gradient intensity is determined as the second direction.
4. The method according to claim 3, wherein the calculating the gradient strength of the calculation block in the candidate direction according to the pixel value of the pixel point included in each of the calculation blocks comprises:

   For any one of the calculation blocks, according to the alternative direction, n pairs of pixel point pairs are selected from the calculation block at intervals as reference pixel point pairs;

   Perform differential calculation based on the pixels included in the reference pixel pair;

   The result of the difference calculation is determined as the gradient strength of the calculation block in the candidate direction.
5. The method according to any one of claims 2 to 3, wherein the selecting the candidate direction from the candidate directions according to the relative relationship between the first direction and the second direction, include:

   If the first direction is parallel to the second direction, determining the candidate direction represented by the first direction as the candidate direction;

   If the first direction is perpendicular to the second direction, selecting at least two of the candidate directions as the candidate directions;

   If an included angle of a preset angle is formed between the first direction and the second direction, the candidate direction represented by the first direction and the candidate direction represented by the second direction are determined as the candidate direction; wherein, the preset angle is an angle other than 0 degrees and 90 degrees.
6. The method according to claim 5, wherein the selecting at least two of the candidate directions as the candidate directions comprises:

   All the candidate directions are selected and determined as the candidate directions.
7. The method according to claim 1, wherein the selecting a candidate pattern from patterns provided by a preset coding standard based on patterns around the candidate direction comprises:

   For any of the candidate directions, with the angle pattern corresponding to the candidate direction as the center, according to the first preset step size, select p patterns at equal intervals from the patterns on both sides of the corresponding angle pattern, as the first an alternative mode;

   determining a first candidate mode according to the first candidate mode, and determining a non-angular mode provided by the preset coding standard as a second candidate mode;

   The first candidate mode and the second candidate mode are determined as the candidate modes.
8. The method according to claim 7, wherein the determining the first candidate mode according to the first candidate mode comprises:

   calculating the rate-distortion cost of the current prediction unit in each of the first candidate modes;

   The first candidate mode corresponding to the first q largest rate-distortion costs is used as the second candidate mode;

   An angle mode adjacent to both sides of the second candidate mode is determined as the first candidate mode.
9. The method according to claim 7, wherein after the first candidate mode and the second candidate mode are determined as the candidate modes, the method further comprises:

   The angle representative mode and the first candidate mode are added to the candidate mode.
10. The method according to any one of claims 7-9, wherein after selecting the candidate mode from the modes provided by the preset coding standard according to the mode around the candidate direction, the method further comprises:

    determining a rate-distortion cost corresponding to the candidate mode;

    If the rate-distortion cost corresponding to the candidate mode is greater than a preset rate-distortion cost threshold, the candidate mode is eliminated.
11. The method according to claim 1, wherein the selecting m angle representative modes from angle modes provided by a preset coding standard, comprising:

    Taking the preset angle mode as a starting point, selecting m angle modes at equal intervals from the angle modes provided by the preset coding standard in sequence according to the second preset step size;

    The selected m angle patterns are used as the angle representative patterns.
12. The method according to claim 1 or 11, wherein the first direction corresponding to the current prediction unit is determined from the candidate directions according to coding effect parameters of the current prediction unit in each of the angle representative modes ,include:

    calculating the rate-distortion cost of the current prediction unit in each of the angle representation modes;

    The angle with the smallest rate-distortion cost represents the candidate direction corresponding to the mode, and is determined as the first direction.
13. A mode selection device, characterized in that the device comprises:

    a first selection mode, used for selecting m angle representative modes from the angle modes provided by the preset coding standard; the m is less than the number of angle modes provided by the preset coding standard;

    The first determination mode is used to take the direction represented by the angle representative mode as an alternative direction, and determine the current prediction unit from the alternative directions according to the coding effect parameters of the current prediction unit in each of the angle representative modes The corresponding first direction; the angle corresponding to the first direction represents that the encoding effect parameter in the mode satisfies the first preset condition;

    a second selection mode, configured to select a candidate direction from the candidate directions according to the first direction; the candidate directions include at least the first direction;

    The third selection mode is used to select a candidate mode from modes provided by the preset coding standard based on modes around the candidate direction.
14. A computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the mode selection method according to any one of claims 1 to 12 is implemented.
15. An electronic device, comprising:

    A processor, a memory, and a computer program stored on the memory and executable on the processor, characterized in that, when the processor executes the program, the mode according to any one of claims 1-12 is implemented Method of choosing.
16. A computer program comprising computer readable code which, when run on a computing processing device, causes the computing processing device to perform the mode selection method according to any one of claims 1-12.

Images