CN104702958B - A kind of HEVC inner frame coding methods and system based on spatial coherence - Google Patents

Abstract

The present invention is applied to the communications field there is provided a kind of HEVC inner frame coding methods based on spatial coherence, including：According to the depth and size of current coded unit, select the space in current coded unit adjacent and with an equal amount of adjacent encoder unit, and ask for the depth bounds [a0, a1] of adjacent encoder unit；According to the relation between the complexity of current coded unit and the complexity of adjacent encoder unit and the position of current coded unit, the depth bounds of current coded unit is determined；Current coded unit is encoded according to the depth bounds of current coded unit, and the depth for selecting Least-cost according to prediction cost, as the depth of current coded unit, the coding for terminating current coded unit is simultaneously transferred to next coding unit.Present invention also offers a kind of HEVC intraframe coding systems based on spatial coherence.HEVC inner frame coding methods and system provided by the present invention based on spatial coherence can improve code efficiency.

Description

A HEVC intra-frame coding method and system based on spatial correlation

technical field

The present invention relates to the field of communication, in particular to a spatial correlation-based HEVC intra-frame encoding method and system.

Background technique

With the improvement of living standards, people's requirements for quality of life are also getting higher and higher. Watching video is a major form of entertainment, and people pay more and more attention to the visual enjoyment it brings. As far as the size of the video is concerned, its resolution gradually increases from 176×144 to 4k×2k, and even 8k×4k. Compared with low-resolution video, high-resolution video can provide humans with greater visual enjoyment.

Based on the new-generation video coding standard H.264/AVC, the High Efficiency Video Coding (HEVC) standard adopts more new technologies to improve video compression efficiency, such as advanced motion vector prediction, more Intra-frame prediction mode, asymmetric motion segmentation, sampling adaptive offset, etc. Although these technologies bring a lot of coding gains, they inevitably bring higher coding complexity, which affects the high-efficiency video coding standard HEVC. From theory to practical application.

In order to reduce the complexity of HEVC encoding process in the High Efficiency Video Coding standard, researchers have proposed many effective methods. For example, someone proposes an algorithm for adaptively determining the depth range of a coding unit (CU), someone uses the depths of previously coded frames and adjacent coded CUs to quickly determine the depth range of the current coding unit CU, and someone in The frame layer and the coding unit CU layer respectively propose an algorithm to quickly determine the depth of the coding unit CU to speed up the coding process. Some people use the Bayesian criterion combined with some related features to quickly determine the depth of the coding unit CU and so on.

Although the above methods can remove the redundancy existing in the encoding process to a certain extent, these methods simply use spatial correlation or temporal correlation to quickly determine the depth range of the coding unit CU, but do not take into account that there are often In the case where the coding unit CU is not strongly correlated with the spatio-temporal adjacent coding unit CU, the prediction of the depth range of the coding unit CU will be inaccurate, resulting in different degrees of coding efficiency reduction.

Therefore, it is urgent to design a new HEVC intra-frame encoding method and system based on spatial correlation, so as to improve encoding efficiency.

Contents of the invention

In view of this, the object of the present invention is to provide a HEVC intra-frame encoding method and system based on spatial correlation, aiming at solving the problem of low encoding efficiency in the prior art.

The present invention is achieved in this way, a HEVC intra-frame encoding method based on spatial correlation, comprising:

According to the depth and size of the current coding unit, select adjacent coding units that are spatially adjacent to the current coding unit and have the same size, and obtain the depth range [a ₀ , a ₁ ] of the adjacent coding unit;

determining the depth range of the current coding unit according to the relationship between the complexity of the current coding unit and the complexity of the neighboring coding units and the location of the current coding unit;

Encode the current coding unit according to the depth range of the current coding unit, and select the depth with the least cost as the depth of the current coding unit according to the prediction cost, end the coding of the current coding unit and transfer to the next coding unit.

Preferably, the step of encoding the current coding unit according to the depth range of the current coding unit, and selecting the depth with the least cost as the depth of the current coding unit according to the prediction cost, ending the coding of the current coding unit and turning to the next coding unit include:

judging whether the depth of the current coding unit is within the determined depth range of the current coding unit;

If the depth of the current coding unit is not within the determined depth range of the current coding unit, determine whether the depth of the current coding unit is greater than or equal to the upper limit of the determined depth range of the current coding unit;

If the depth of the current coding unit is greater than or equal to the determined upper limit of the depth range of the current coding unit, end the splitting process of the current coding unit;

judging whether the current coding unit is the last coding unit to be searched in the pruned coding tree unit;

If the current coding unit is the last coding unit to be searched in the pruned coding tree unit, the recursive division process of the current coding tree unit is ended, and the coding of the coding unit corresponding to the current coding tree unit is completed.

Preferably, after the step of judging whether the depth of the current coding unit is within the determined depth range of the current coding unit, the method further includes:

If the depth of the current coding unit is within the determined depth range of the current coding unit, perform an intra prediction process, calculate the cost corresponding to the best prediction mode, and select the mode with the smallest cost as the best mode.

Preferably, the step of determining the depth range of the current coding unit according to the relationship between the complexity of the current coding unit and the complexity of the adjacent coding units and the location of the current coding unit is specifically include:

If the current coding unit belongs to the topmost or leftmost coding unit, determine that the original depth range of the current coding unit is [0, 3];

Or, if the complexity of the current coding unit is greater than the maximum value of the complexity of the adjacent coding unit, determine that the depth range of the current coding unit is [a ₀ , min(a ₁ +N, 3)] , N∈[0,3];

Or, if the complexity of the current coding unit is less than the minimum value of the complexity of the adjacent coding units, determine that the depth range of the current coding unit is [max(0, a ₀ -N), a ₁ ] , N∈[0,3];

Or, if the complexity of the current coding unit is between the complexities of the adjacent coding units, determine that the depth range of the current coding unit is [a ₀ , a ₁ ].

Preferably, according to the depth and size of the current coding unit, select adjacent coding units that are spatially adjacent to the current coding unit and have the same size, and calculate the depth range of the adjacent coding units [a ₀ , a ₁ ] steps include:

According to the depth and size of the current coding unit, select the adjacent coding units adjacent to the left and above of the current coding unit and have the same size, and calculate the union of the depth ranges of the adjacent coding units on the left and above to obtain the depth range [a ₀ , a ₁ ] of the adjacent coding unit.

On the other hand, the present invention also provides a HEVC intra-frame encoding system based on spatial correlation, the system comprising:

The selection module is configured to select, according to the depth and size of the current coding unit, adjacent coding units that are spatially adjacent to the current coding unit and have the same size, and calculate the depth range of the adjacent coding units [a ₀ , a ₁ ];

A range determination module, configured to determine the depth range of the current coding unit according to the relationship between the complexity of the current coding unit and the complexity of the adjacent coding units and the location of the current coding unit;

The encoding prediction module is used to encode the current coding unit according to the depth range of the current coding unit, and select the depth with the lowest cost as the depth of the current coding unit according to the prediction cost, end the coding of the current coding unit and transfer to the next coding unit.

Preferably, the coding prediction module includes:

A first judging submodule, configured to judge whether the depth of the current coding unit is within the determined depth range of the current coding unit;

The second judging submodule is configured to judge whether the depth of the current coding unit is greater than or equal to the determined depth of the current coding unit if the depth of the current coding unit is not within the determined depth range of the current coding unit. the upper limit of the depth range of the coding unit;

Ending the splitting submodule, configured to end the splitting process of the current coding unit if the depth of the current coding unit is greater than or equal to the determined upper limit of the depth range of the current coding unit;

The third judging submodule is used to judge whether the current coding unit is the last coding unit to be searched in the pruned coding tree unit;

Wherein, the end segmentation sub-module is also used to end the recursive segmentation process of the current coding tree unit if the current coding unit is the last coding unit to be searched in the pruned coding tree unit, and complete the corresponding The encoding of the coding unit.

Preferably, the coding prediction module also includes:

The execution sub-module is used to execute the intra prediction process if the depth of the current coding unit is within the determined depth range of the current coding unit, and calculate the cost corresponding to the best prediction mode, and select the mode with the smallest cost is the best mode.

Preferably, the range determination module is specifically used for:

If the current coding unit belongs to the topmost or leftmost coding unit, determine that the original depth range of the current coding unit is [0, 3];

Or, if the complexity of the current coding unit is greater than the maximum value of the complexity of the adjacent coding unit, determine that the depth range of the current coding unit is [a ₀ , min(a ₁ +N, 3)] , N∈[0,3];

Or, if the complexity of the current coding unit is less than the minimum value of the complexity of the adjacent coding units, determine that the depth range of the current coding unit is [max(0, a ₀ -N), a ₁ ] , N∈[0,3];

Or, if the complexity of the current coding unit is between the complexities of the adjacent coding units, determine that the depth range of the current coding unit is [a ₀ , a ₁ ].

Preferably, the selection module is specifically configured to: select, according to the depth and size of the current coding unit, an adjacent coding unit that is adjacent to the left and above the current coding unit and has the same size, and calculates the left and upper The depth range [a0, a1] of the adjacent CU is obtained by the union of the depth ranges of the adjacent CUs above.

In the embodiment of the present invention, in the technical solution provided by the present invention, the intra-frame coding process of each frame of image in the video sequence is processed according to the coding tree unit (Coding Tree Unit, CTU), and the coding tree unit is processed by depth-first recursive method. By traversing all the coding units (Coding Unit, CU) in the coding tree unit, the present invention accurately determines the depth range of the coding unit in a recursive manner according to the spatial correlation of the video sequence, and reduces the CU without basically causing the degradation of the video quality after coding. Traversing the range of the coding tree unit search effectively reduces the complexity of the intra-frame coding process. Using the technical solution provided by the present invention to predict the coded depth of the current coding unit through the coding information of the spatially adjacent coding units, construct a more effective depth range for the coding unit, and reduce the coding of rate-distortion optimization (RDO) The number of units, so as to reduce the complexity of the intra-frame encoding process more effectively and improve the encoding speed under the premise of ensuring the encoding performance.

Description of drawings

FIG. 1 is a flowchart of an HEVC intra-frame encoding method based on spatial correlation in an embodiment of the present invention;

Fig. 2 is a detailed method flowchart of step S13 shown in Fig. 1 in an embodiment of the present invention;

3 is a schematic diagram of recursive partitioning of a coding tree unit CTU in an embodiment of the present invention;

FIG. 4 is a quadtree segmentation structure finally selected by the coding tree unit CTU in an embodiment of the present invention;

5 is a schematic structural diagram of an HEVC intra-coding system based on spatial correlation in an embodiment of the present invention;

FIG. 6 is a schematic diagram of the internal structure of the encoding prediction module 103 shown in FIG. 5 in an embodiment of the present invention;

Figure 7a and Figure 7b are comparative test charts of the three encoding methods.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

The specific embodiment of the present invention provides a HEVC intra-frame encoding method based on spatial correlation, which mainly includes the following steps:

S11. According to the depth and size of the current coding unit, select adjacent coding units that are spatially adjacent to the current coding unit and have the same size, and calculate the depth range [a ₀ , a ₁ ] of the adjacent coding units ];

S12. Determine the depth range of the current coding unit according to the relationship between the complexity of the current coding unit and the complexity of the adjacent coding units and the location of the current coding unit;

S13. Encode the current coding unit according to the depth range of the current coding unit, and select the depth with the lowest cost according to the prediction cost as the depth of the current coding unit, end the coding of the current coding unit and transfer to the next coding unit.

In the HEVC intra-frame coding method based on spatial correlation provided by the present invention, the intra-frame coding process of each frame of image in the video sequence is processed by coding tree unit (Coding Tree Unit, CTU), and the coding tree unit is processed by depth The preferred recursive method traverses all the coding units (Coding Unit, CU) in the coding tree unit. According to the spatial correlation of the video sequence, the present invention uses a recursive method to accurately determine the depth range of the coding unit, which basically does not cause the video quality to decline after coding. On the premise of reducing the range of traversing the coding tree unit search, the complexity of the intra coding process is effectively reduced. Using the technical solution provided by the present invention to predict the coded depth of the current coding unit through the coding information of the spatially adjacent coding units, construct a more effective depth range for the coding unit, and reduce the coding units that perform rate-distortion optimization (Rate-distortion Optimization, RDO) Number, so that under the premise of ensuring the coding performance, the complexity of the intra-frame coding process can be more effectively reduced, and the coding speed can be improved.

A spatial correlation-based HEVC intra-frame encoding method provided by the present invention will be described in detail below.

Please refer to FIG. 1 , which is a flowchart of an HEVC intra-frame encoding method based on spatial correlation in an embodiment of the present invention.

In step S11, according to the depth and size of the current coding unit, select adjacent coding units that are spatially adjacent to the current coding unit and have the same size, and calculate the depth range [a ₀ , a ₁ ].

In this embodiment, denoted a ₀ represents the lower limit of the depth range union of adjacent coding units CU that are spatially adjacent to the current coding unit and have the same size, and a ₁ represents adjacent coding units that are spatially adjacent and have the same size The upper bound of the union of unit CU depth ranges.

In this embodiment, the coding unit (Coding Unit, CU) is the basic unit of coding, similar to the concept of macroblock in the new generation video coding standard H.264/AVC, it is a coding tree recursively divided by quadtree structure The leaf node of the unit (Coding Tree Unit, CTU). The largest coding unit (Largest Coding Unit, LCU) size and the largest split depth of the coding tree unit CTU can be specified in the sequence parameter set (Sequence Parameter Set, SPS).

In this embodiment, the coding tree unit CTU uses the split flag to determine whether the coding unit CU needs to be further split. When splitting is required, the coding unit CU can be divided into four lower-level CUs of the same size. The prediction unit PU (PredictionUnit) is the basic unit of prediction, which determines which partition type the coding unit CU adopts for prediction, and its root node is at the coding unit CU layer. There are two types of division in the intra prediction mode: non-division and quartering, and the quartering mode is executed only when the recursive division of the coding unit CU reaches the minimum size. The transformation unit TU (Transform Unit) is the basic unit of residual transformation, and the root node is also at the coding unit CU layer. The processing of the transform unit TU also adopts a quadtree structure, and its quadtree is called a residual quadtree (RQT). In the sequence parameter set SPS, the maximum transform block (Transform Block, TB) size and the maximum segmentation depth of the residual quadtree RQT can be specified. The transform block TB basically adopts discrete cosine transform (Discrete Cosine Transform, DCT), but if the transform block TB has a size of 4×4 and intra-frame prediction, discrete sine transform (Discrete Sine Transform, DST) is selected.

In step S12, the depth range of the current coding unit is determined according to the relationship between the complexity of the current coding unit and the complexity of the adjacent coding units and the location of the current coding unit.

In this embodiment, the depth range of the current coding unit is determined according to the relationship between the complexity of the current coding unit and the complexity of the adjacent coding units and the location of the current coding unit The steps specifically include:

If the current coding unit belongs to the topmost or leftmost coding unit, determine that the original depth range of the current coding unit is [0, 3];

Or, if the complexity of the current coding unit is greater than the maximum value of the complexity of the adjacent coding unit, determine that the depth range of the current coding unit is [a ₀ , min(a ₁ +N, 3)] , N∈[0,3]; N is set to 1 in this embodiment;

Or, if the complexity of the current coding unit is less than the minimum value of the complexity of the adjacent coding units, determine that the depth range of the current coding unit is [max(0, a ₀ -N), a ₁ ] , N∈[0,3]; N is set to 1 in this embodiment;

Or, if the complexity of the current coding unit is between the complexities of the adjacent coding units, determine that the depth range of the current coding unit is [a ₀ , a ₁ ].

In step S13, encode the current coding unit according to the depth range of the current coding unit, and select the depth with the lowest cost as the depth of the current coding unit according to the prediction cost, end the coding of the current coding unit and transfer to the next coding unit.

In this embodiment, step S13 specifically includes steps S131-S138, as shown in FIG. 2 .

Please refer to FIG. 2 , which is a detailed method flowchart of step S13 shown in FIG. 1 in an embodiment of the present invention.

In step S131, it is judged whether the depth of the current coding unit is within the determined depth range of the current coding unit.

If the depth of the current coding unit is not within the determined depth range of the current coding unit, then in step S132, it is judged whether the depth of the current coding unit is greater than or equal to the determined depth of the current coding unit The upper limit of the depth range.

In this embodiment, if the depth of the current coding unit is within the determined depth range of the current coding unit, then in step S136, perform an intra prediction process, and calculate the cost corresponding to the best prediction mode, Choose the mode with the least cost as the best mode.

In this embodiment, during intra-frame prediction, the coding tree unit CTU is recursively divided according to the quadtree, and each leaf node constitutes a coding unit CU. The coding unit CU includes a luminance coding block (Coding Block, CB), and two chroma A coding block CB, and corresponding syntax elements. In the high-efficiency video coding standard HEVC, the coding tree unit CTU will be recursively divided according to the maximum coding unit size and the maximum recursion depth specified in the sequence parameter set SPS. In order to describe the process of recursive partitioning in detail, assuming that the maximum coding unit size specified in the sequence parameter set SPS is 32×32 and the maximum recursion depth is 3, then the corresponding recursive partitioning diagram of the coding tree unit CTU is shown in Figure 3.

Please refer to FIG. 3 , which is a schematic diagram of recursive partitioning of a coding tree unit CTU in an embodiment of the present invention. In Figure 3, the numbers 1-25 indicate the order of processing, first execute 1, that is, find the cost corresponding to the best prediction mode for a 32×32 block, and use two bits occupied by data such as image distortion and coding residual The parameters comprehensively measure the cost corresponding to the coding unit CU, and the specific cost function is shown in Equation 1. Since the coding tree unit CTU adopts a depth-first search method, 2 is executed after 1 is executed, that is, the cost corresponding to the best prediction mode of the coding unit CU numbered 2 is calculated. However, the coding unit CU numbered 3 is processed in the same way. Since the coding unit CU numbered 3 has reached the maximum depth specified in the sequence parameter set SPS, the coding unit CU numbered 3 does not need to be divided after processing, but the same layer numbers 4, 5, and 6 are processed sequentially. The coding unit CU of . The above is the division process of the coding tree unit CTU. In order to judge whether the coding unit CU numbered 2 uses 16×16 block coding or four 8×8 block coding in the final actual coding process, it is necessary to perform the operation numbered 7, that is, compare the coding units numbered 3-6 The sum of the cost corresponding to the CU and the cost of the coding unit CU numbered 2 is used to encode the coding unit CU numbered 2 in a manner corresponding to the minimum cost. This is the pruning process of the coding unit. The segmentation and pruning process of the coding unit CU is performed sequentially until the processing of the entire coding tree unit CTU is completed. The process described above is to select a combination with the smallest corresponding cost among different CU combinations as the best coding mode coding tree unit CTU, and this process is called Rate Distortion Optimization (RDO). Obviously, the rate-distortion optimization RDO process selects the best coding method for the coding tree unit CTU by traversing all the coding unit CU combinations. Although this method has high coding efficiency, the coding complexity is also the highest, and when the largest coding unit When the CU size is constant, the coding complexity increases rapidly with the increase of the maximum coding depth set in the sequence parameter set.

RDCost=D+λ·R (1)

Among them, RDCost represents the cost of image distortion and bit synthesis required for encoding after the current block is encoded, D represents the difference between the reconstructed pixel and the original pixel, R represents the bit required to encode the current block information, λ represents the Lagrangian multiplication son.

From the above analysis, it can be seen that the high-efficiency video coding standard HEVC traverses all the leaf nodes in the coding tree unit CTU according to the parameter range set in the sequence parameter set SPS to find the optimal coding unit CU combination. However, the final selected coding unit CU is related to the video content. The more complicated the content, the smaller the selected coding unit CU, and the flatter the content, the larger the selected coding unit CU. Therefore, if the same coded depth range is used for all coding units CU without considering the characteristics of the video content, a large coding redundancy will be generated. To illustrate this issue in more detail, we illustrate with an example. Fig. 4 is a quadtree partition structure finally selected by the coding tree unit CTU.

As shown in Figure 4, when the largest coding unit (Largest Coding Unit, LCU) set in the sequence parameter set SPS is 64×64, and the maximum segmentation depth is 4, the possible coding unit CU sizes are 64×64, 32×32 , 16×16, 8×8, and the corresponding coding depths are 0, 1, 2, 3 respectively. In the encoding process, the coding tree unit CTU traverses all the depths of the coding unit CU through recursive partitioning, so as to select the best coding unit CU depth to code the coding tree unit CTU. If the coding tree unit CTU uses a fixed depth range, the coding tree unit CTU will be recursively split from depth 0 to depth 3 step by step. Since the depth of the finally selected coding unit CU is related to the video content, the depth of the finally selected coding unit CU may not all include depth 0, depth 1, depth 2, and depth 3. In Figure 3, the depth of the finally selected coding unit CU does not include depth 0, but only depth 1, depth 2, and depth 3, so the depth range of the current coding tree unit CTU is [1,3]. In the actual encoding process, the high-efficiency video coding standard HEVC will start from depth 0 and recursively split to depth 3 step by step. Therefore, all operations related to depth 0 have no practical significance for the current coding tree unit CTU coding, which is the redundancy in the coding process. During the encoding process, if the depth range of the current coding tree unit CTU can be accurately predicted in advance, unnecessary depths can be directly skipped, thereby greatly reducing the encoding complexity without significantly reducing the quality of the encoded video.

Please continue to refer to FIG. 2. In this embodiment, if the depth of the current coding unit is greater than or equal to the determined upper limit of the depth range of the current coding unit, then in step S133, end the coding of the current coding unit. Segmentation process.

In this embodiment, if the depth of the current coding unit is smaller than the determined upper limit of the depth range of the current coding unit, then in step S137, add 1 to the depth of the current coding unit. In this embodiment, after adding 1 to the depth of the current coding unit, continue to execute step S11 shown in FIG. 1 .

In step S134, it is judged whether the current coding unit is the last coding unit to be searched in the pruned coding tree unit.

If the current coding unit is the last coding unit to be searched in the pruned coding tree unit, then in step S135, the recursive division process of the current coding tree unit is ended, and the coding of the coding unit corresponding to the current coding tree unit is completed.

In this embodiment, if the current coding unit is not the last coding unit to be searched in the pruned coding tree unit, turn to the next coding unit and determine the next coding unit as the current coding unit, as in step As shown in S138, the following steps are then repeatedly executed from step S11 shown in FIG. 1 , and will not be described again here.

In the HEVC intra-frame coding method based on spatial correlation provided by the present invention, the intra-frame coding process of each frame of image in the video sequence is processed by coding tree unit (Coding Tree Unit, CTU), and the coding tree unit is processed by depth The preferred recursive method traverses all the coding units (Coding Unit, CU) in the coding tree unit. According to the spatial correlation of the video sequence, the present invention uses a recursive method to accurately determine the depth range of the coding unit, which basically does not cause the video quality to decline after coding. On the premise of reducing the range of traversing the coding tree unit search, the complexity of the intra coding process is effectively reduced. Using the technical solution provided by the present invention to predict the coded depth of the current coding unit through the coding information of the spatially adjacent coding units, construct a more effective depth range for the coding unit, and reduce the coding units that perform rate-distortion optimization (Rate-distortion Optimization, RDO) Number, so that under the premise of ensuring the coding performance, the complexity of the intra-frame coding process can be more effectively reduced, and the coding speed can be improved.

The specific embodiment of the present invention also provides a HEVC intra coding system 10 based on spatial correlation, which mainly includes:

The selection module 101 is configured to select, according to the depth and size of the current coding unit, adjacent coding units that are spatially adjacent to the current coding unit and have the same size, and calculate the depth range of the adjacent coding units [a ₀ , a ₁ ];

A range determination module 102, configured to determine the depth range of the current coding unit according to the relationship between the complexity of the current coding unit and the complexity of the adjacent coding units and the location of the current coding unit;

The encoding prediction module 103 is configured to encode the current coding unit according to the depth range of the current coding unit, and select the depth with the lowest cost according to the prediction cost as the depth of the current coding unit, end the coding of the current coding unit and transfer to the next coding unit .

In the HEVC intra-frame encoding system 10 based on spatial correlation provided by the present invention, the intra-frame encoding process of each frame of image in the video sequence is processed according to the coding tree unit (Coding Tree Unit, CTU), and the coding tree unit is passed through The depth-first recursive method traverses all the coding units (Coding Unit, CU) in the coding tree unit. According to the spatial correlation of the video sequence, the present invention uses the recursive method to accurately determine the depth range of the coding unit, which basically does not affect the video quality after coding. On the premise of descending, the range of traversing the coding tree unit search is reduced, which effectively reduces the complexity of the intra coding process. Using the technical solution provided by the present invention to predict the coded depth of the current coding unit through the coding information of the spatially adjacent coding units, construct a more effective depth range for the coding unit, and reduce the coding units that perform rate-distortion optimization (Rate-distortion Optimization, RDO) Number, so that under the premise of ensuring the coding performance, the complexity of the intra-frame coding process can be more effectively reduced, and the coding speed can be improved.

A spatial correlation-based HEVC intra-coding system 10 provided by the present invention will be described in detail below.

Please refer to FIG. 5 , which is a schematic structural diagram of an HEVC intra-coding system 10 based on spatial correlation in an embodiment of the present invention. In this embodiment, the HEVC intra coding system 10 based on spatial correlation includes a selection module 101 , a range determination module 102 and a coding prediction module 103 .

The selection module 101 is configured to select, according to the depth and size of the current coding unit, adjacent coding units that are spatially adjacent to the current coding unit and have the same size, and calculate the depth range of the adjacent coding units [a ₀ ,a ₁ ].

In this embodiment, denoted a ₀ represents the lower limit of the depth range union of adjacent coding units CU that are spatially adjacent to the current coding unit and have the same size, and a ₁ represents adjacent coding units that are spatially adjacent and have the same size The upper bound of the union of unit CU depth ranges.

In this embodiment, the coding unit (Coding Unit, CU) is the basic unit of coding, similar to the concept of macroblock in the new generation video coding standard H.264/AVC, it is a coding tree recursively divided by quadtree structure The leaf node of the unit (Coding Tree Unit, CTU). The largest coding unit (Largest Coding Unit, LCU) size and the largest split depth of the coding tree unit CTU can be specified in the sequence parameter set (Sequence Parameter Set, SPS).

In this embodiment, the coding tree unit CTU uses the split flag to determine whether the coding unit CU needs to be further split. When splitting is required, the coding unit CU can be divided into four lower-level CUs of the same size. The prediction unit PU (PredictionUnit) is the basic unit of prediction, which determines which partition type the coding unit CU adopts for prediction, and its root node is at the coding unit CU layer. There are two types of division in the intra prediction mode: non-division and quartering, and the quartering mode is executed only when the recursive division of the coding unit CU reaches the minimum size. The transformation unit TU (Transform Unit) is the basic unit of residual transformation, and the root node is also at the coding unit CU layer. The processing of the transform unit TU also adopts a quadtree structure, and its quadtree is called a residual quadtree (RQT). In the sequence parameter set SPS, the maximum transform block (Transform Block, TB) size and the maximum segmentation depth of the residual quadtree RQT can be specified. The transform block TB basically adopts discrete cosine transform (Discrete Cosine Transform, DCT), but if the transform block TB has a size of 4×4 and intra-frame prediction, discrete sine transform (Discrete Sine Transform, DST) is selected.

The range determination module 102 is configured to determine the depth range of the current coding unit according to the relationship between the complexity of the current coding unit and the complexity of the adjacent coding units and the location of the current coding unit.

In this embodiment, the range determination module 102 is specifically used for:

If the current coding unit belongs to the topmost or leftmost coding unit, determine that the original depth range of the current coding unit is [0, 3];

Or, if the complexity of the current coding unit is greater than the maximum value of the complexity of the adjacent coding unit, determine that the depth range of the current coding unit is [a ₀ , min(a ₁ +N, 3)] , N∈[0,3];

Or, if the complexity of the current coding unit is less than the minimum value of the complexity of the adjacent coding units, determine that the depth range of the current coding unit is [max(0, a ₀ -N), a ₁ ] , N∈[0,3];

Or, if the complexity of the current coding unit is between the complexities of the adjacent coding units, determine that the depth range of the current coding unit is [a ₀ , a ₁ ].

The encoding prediction module 103 is configured to encode the current coding unit according to the depth range of the current coding unit, and select the depth with the lowest cost according to the prediction cost as the depth of the current coding unit, end the coding of the current coding unit and transfer to the next coding unit .

In this embodiment, the encoding prediction module 103 specifically includes a first judgment submodule 1031, a second judgment submodule 1032, an end segmentation submodule 1033, a third judgment submodule 1034, an execution submodule 1035, and a depth increase submodule 1036, As shown in Figure 6.

Please refer to FIG. 6 , which is a schematic diagram of the internal structure of the coding prediction module 103 shown in FIG. 5 in an embodiment of the present invention.

The first judging sub-module 1031 is configured to judge whether the depth of the current coding unit is within the determined depth range of the current coding unit.

The second judging submodule 1032 is configured to judge whether the depth of the current coding unit is greater than or equal to the determined depth range of the current coding unit if the depth of the current coding unit is not within the determined depth range of the current coding unit. The upper bound of the depth range for the current coding unit.

Executing sub-module 1035, configured to perform an intra-frame prediction process if the depth of the current coding unit is within the determined depth range of the current coding unit, and calculate the cost corresponding to the best prediction mode, and select the one with the smallest cost mode is the best mode. In this embodiment, the specific method steps for performing the intra prediction process and calculating the cost corresponding to the best prediction mode are as described in the above-mentioned step S136, and will not be repeated here. Please refer to the above-mentioned step S136 and the corresponding FIG. 2 with Figure 3.

End splitting sub-module 1033, configured to end the splitting process of the current coding unit if the depth of the current coding unit is greater than or equal to the determined upper limit of the depth range of the current coding unit.

The depth increasing sub-module 1036 is configured to add 1 to the depth of the current coding unit if the depth of the current coding unit is smaller than the determined upper limit of the depth range of the current coding unit. In this embodiment, after adding 1 to the depth of the current coding unit, the selection module 101 is further configured to continue to select the spatially adjacent current coding unit and have the same size of adjacent coding units, and obtain the depth range [a ₀ , a ₁ ] of the adjacent coding units.

The third judging sub-module 1034 is configured to judge whether the current coding unit is the last coding unit to be searched in the pruned coding tree units.

Wherein, the end segmentation sub-module 1033 is also used to end the recursive segmentation process of the current coding tree unit and complete the current coding tree unit if the current coding unit is the last coding unit to be searched in the pruned coding tree unit The encoding of the corresponding coding unit.

In this embodiment, the selection module 101 is further configured to switch to the next coding unit and determine the next coding unit if the current coding unit is not the last coding unit to be searched in the pruned coding tree unit is the current code unit.

In the HEVC intra-frame encoding system 10 based on spatial correlation provided by the present invention, the intra-frame encoding process of each frame of image in the video sequence is processed according to the coding tree unit (Coding Tree Unit, CTU), and the coding tree unit is passed through The depth-first recursive method traverses all the coding units (Coding Unit, CU) in the coding tree unit. According to the spatial correlation of the video sequence, the present invention uses the recursive method to accurately determine the depth range of the coding unit, which basically does not affect the video quality after coding. On the premise of descending, the range of traversing the coding tree unit search is reduced, which effectively reduces the complexity of the intra coding process. Using the technical solution provided by the present invention to predict the coded depth of the current coding unit through the coding information of the spatially adjacent coding units, construct a more effective depth range for the coding unit, and reduce the coding units that perform rate-distortion optimization (Rate-distortion Optimization, RDO) Number, so that under the premise of ensuring the coding performance, the complexity of the intra-frame coding process can be more effectively reduced, and the coding speed can be improved.

Compared with the prior art, the technical solution provided by the present invention has the advantage of improving the encoding speed, for example, reference [4] (X.Li, J.An, X.Guo, S.Lei, "Adaptive CU Depth Range," JCTVC -E090, JCTVC of ISO/IEC and ITU-T, Geneva, Switzerland, Apr.2011.) proposes an algorithm for adaptively determining the depth range of the coding unit CU to reduce coding complexity. Although the above reference [4] method can remove the redundancy existing in the encoding process to a certain extent, this method only simply uses spatial correlation or temporal correlation to quickly determine the depth range of the coding unit CU, but does not take into account the In the frame image, there is often a situation where the CU is not strongly correlated with the adjacent CU in time and space, which will result in inaccurate prediction of the depth range of the CU, resulting in varying degrees of reduction in coding efficiency.

To solve the problem in the prior art that the complexity of the intra coding process cannot be sufficiently reduced, the present invention uses the depth range of the adjacent coded coding unit CU, and the complexity of the current coding unit CU and the adjacent coded coding The relationship between unit CU complexity accurately predicts the depth range of the current coding unit CU. Under the premise of ensuring that the video quality after encoding is basically not reduced, unnecessary depths are skipped, thereby greatly reducing the complexity of the high-efficiency video coding standard HEVC intra-frame coding process Spend. It should be pointed out that the complexity of the coding unit CU in the present invention is measured by the sum of absolute differences.

The comparative experimental schemes include the original high-efficiency video coding standard HEVC coding platform scheme, the fast intra-frame coding scheme of reference [4] and the method of the present invention. The coding experiment results show that: in terms of compression efficiency, under the same code rate, the peak signal-to-noise ratio of the method of the present invention decreases to an average of 0.03dB compared to the original high-efficiency video coding standard HEVC coding platform, which can be ignored. The comparison results are shown in Figure 7a and Figure 7b shown.

Please refer to Fig. 7a and Fig. 7b, which are comparative test charts of the three encoding methods. The specific three coding methods include: the method of the present invention, the HEVC coding platform of the original high-efficiency video coding standard, and the reference [4] method under the two configurations of full intra-frame high-efficiency profile and full-frame main profile. Comparison of partial sequence rate-distortion performance .

The speed of the intra-frame coding process is increased by 27.99% to 67.08% (average 40.16%) under the full intra-frame high-efficiency profile configuration, and the complexity is reduced by 21.13% compared with the existing method reference [4]; compared with the original high-efficiency video coding standard HEVC coding platform, under the full intra-frame main profile configuration, the speed of the intra-frame coding process is increased by 28.03% to 63.89% (average 39.41%), and the complexity is further reduced by 20.73% compared with the existing method reference [4]. In addition, the present invention mainly solves the problem of a large amount of redundancy in the fixed depth range of the coding unit CU, and can be combined with other traditional intra-frame fast methods to further reduce the complexity of intra-frame coding.

In this embodiment, the experimental results of the three encoding methods are shown in Table 1 below, and the average peak signal-to-noise ratio increase BD-PSNR (dB), the average bit rate increase BD-BitRate (%), and the time saving Time -Saving(TS)(%) evaluates the performance of the proposed algorithm.

Table 1

Among them, the definition of time saving TS is shown in the following formula:

Among them, T _HM8.0 (QP _i ) represents the total coding time of coding platform HM8.0 when QP=QP _i ; T _proposed (QP _i ) represents the total time of proposing method coding when QP=QP _i .

In the embodiment of the present invention, in the technical solution provided by the present invention, the intra-frame coding process of each frame of image in the video sequence is processed according to the coding tree unit (Coding Tree Unit, CTU), and the coding tree unit is processed by depth-first recursive method. By traversing all the coding units (Coding Unit, CU) in the coding tree unit, the present invention accurately determines the depth range of the coding unit in a recursive manner according to the spatial correlation of the video sequence, and reduces the CU without basically causing the degradation of the video quality after coding. Traversing the range of the coding tree unit search effectively reduces the complexity of the intra-frame coding process. Using the technical solution provided by the present invention to predict the coded depth of the current coding unit through the coding information of the spatially adjacent coding units, construct a more effective depth range for the coding unit, and reduce the coding of rate-distortion optimization (RDO) The number of units, so as to reduce the complexity of the intra-frame encoding process more effectively and improve the encoding speed under the premise of ensuring the encoding performance.

It is worth noting that in the above embodiments, the units included are only divided according to the functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are also It is only for the convenience of distinguishing each other, and is not intended to limit the protection scope of the present invention.

In addition, those of ordinary skill in the art can understand that all or part of the steps in the methods of the above-mentioned embodiments can be completed by instructing related hardware through programs, and the corresponding programs can be stored in a computer-readable storage medium. Storage media, such as ROM/RAM, magnetic disk or optical disk, etc.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. within range.

Claims (10)

1. a kind of HEVC inner frame coding methods based on spatial coherence, it is characterised in that methods described includes：

According to the depth and size of current coded unit, select the space in the current coded unit adjacent and with same The adjacent encoder unit of size, and ask for the depth bounds [a of the adjacent encoder unit₀, a₁]；

According to the relation between the complexity of the current coded unit and the complexity of the adjacent encoder unit and described The position of current coded unit, determines the depth bounds of the current coded unit；

Current coded unit is encoded according to the depth bounds of current coded unit, and according to prediction cost selection cost most Small depth terminates the coding of current coded unit and is transferred to next coding unit as the depth of current coded unit.

2. the HEVC inner frame coding methods as claimed in claim 1 based on spatial coherence, it is characterised in that the basis is worked as The depth bounds of preceding coding unit is encoded to current coded unit, and is made according to the depth of prediction cost selection Least-cost For the depth of current coded unit, terminate the coding of current coded unit and include the step of being transferred to next coding unit：

Judge the depth of the current coded unit whether within the depth bounds for the current coded unit having determined；

If the depth of the current coded unit within the depth bounds for the current coded unit having determined, is not sentenced Whether the depth of the disconnected current coded unit is more than or equal to the depth bounds for the current coded unit having determined The upper limit；

If the depth of the current coded unit is more than or equal to the depth bounds for the current coded unit having determined The upper limit, then terminate the cutting procedure of the current coded unit；

Whether judge the current coded unit is last coding unit to be searched in code tree unit after trimming；

If the current coded unit is last coding unit to be searched in code tree unit after trimming, terminate current The recursive subdivision process of code tree unit, completes the coding of the corresponding coding unit of present encoding tree unit.

3. the HEVC inner frame coding methods as claimed in claim 2 based on spatial coherence, it is characterised in that judge described The step of whether depth of the current coded unit within the depth bounds for the current coded unit having determined it Afterwards, methods described also includes：

If the depth of the current coded unit is performed within the depth bounds for the current coded unit having determined Intra-prediction process, and optimum prediction mode correspondence cost is calculated, the pattern of selection Least-cost is optimal mode.

4. the HEVC inner frame coding methods as claimed in claim 3 based on spatial coherence, it is characterised in that described according to institute State the relation and the present encoding list between the complexity of current coded unit and the complexity of the adjacent encoder unit The step of position of member, depth bounds for determining the current coded unit, specifically includes：

If the current coded unit belongs to the coding unit at the top or leftmost edge, it is determined that the present encoding list The original depth bounds of member is [0,3]；

Or, if the complexity of the current coded unit is more than the maximum of the complexity of the adjacent encoder unit, really The depth bounds of the fixed current coded unit is [a₀, min (a₁+ N, 3)], N ∈ [0,3]；

Or, if the complexity of the current coded unit is less than the minimum value of the complexity of the adjacent encoder unit, really The depth bounds of the fixed current coded unit is [max (0, a₀- N), a₁], N ∈ [0,3]；

Or, if the complexity of the current coded unit is between the complexity of the adjacent encoder unit, it is determined that institute The depth bounds for stating current coded unit is [a₀, a₁]。

5. the HEVC inner frame coding methods as claimed in claim 1 based on spatial coherence, it is characterised in that the basis is worked as The depth and size of preceding coding unit, select the space in the current coded unit adjacent and with an equal amount of adjacent Coding unit, and ask for the depth bounds [a of the adjacent encoder unit₀, a₁] the step of include：

According to the depth and size of current coded unit, selection is adjacent with top and have in the left of the current coded unit Have an an equal amount of adjacent encoder unit, and by calculate the union of left and the depth bounds of top adjacent encoder unit come To the depth bounds [a of the adjacent encoder unit₀, a₁]。

6. a kind of HEVC intraframe coding systems based on spatial coherence, it is characterised in that the system includes：

Selecting module, for the depth and size according to current coded unit, selects the space phase in the current coded unit It is adjacent and with an equal amount of adjacent encoder unit, and ask for the depth bounds [a of the adjacent encoder unit₀, a₁]；

Range determination module, for the complexity according to the current coded unit and the adjacent encoder unit complexity it Between relation and the current coded unit position, determine the depth bounds of the current coded unit；

Prediction module is encoded, current coded unit is encoded for the depth bounds according to current coded unit, and according to Predict that the depth of cost selection Least-cost, as the depth of current coded unit, terminates the coding of current coded unit and is transferred to Next coding unit.

7. the HEVC intraframe coding systems as claimed in claim 6 based on spatial coherence, it is characterised in that the coding is pre- Surveying module includes：

First judging submodule, for judging the depth of the current coded unit whether in the present encoding having determined Within the depth bounds of unit；

Second judging submodule, if the depth for the current coded unit is not in the current coded unit having determined Depth bounds within, then judge whether the depth of the current coded unit is more than or equal to have determined described current The upper limit of the depth bounds of coding unit；

Terminate segmentation submodule, if the depth for the current coded unit is described current more than or equal to what is had determined The upper limit of the depth bounds of coding unit, then terminate the cutting procedure of the current coded unit；

3rd judging submodule, for judge the current coded unit whether be after trimming in code tree unit it is to be searched most Latter coding unit；

Wherein, it is described to terminate segmentation submodule, if it is to wait to search in code tree unit after trimming to be additionally operable to the current coded unit Last coding unit of rope, then terminate the recursive subdivision process of present encoding tree unit, completes present encoding tree unit pair The coding for the coding unit answered.

8. the HEVC intraframe coding systems as claimed in claim 7 based on spatial coherence, it is characterised in that the coding is pre- Surveying module also includes：

Implementation sub-module, if for the current coded unit depth the current coded unit having determined depth Within the scope of, then intra-prediction process is performed, and optimum prediction mode correspondence cost is calculated, the pattern of selection Least-cost is most Good pattern.

9. the HEVC intraframe coding systems as claimed in claim 8 based on spatial coherence, it is characterised in that the scope is true Cover half block, specifically for：

If the current coded unit belongs to the coding unit at the top or leftmost edge, it is determined that the present encoding list The original depth bounds of member is [0,3]；

Or, if the complexity of the current coded unit is more than the maximum of the complexity of the adjacent encoder unit, really The depth bounds of the fixed current coded unit is [a₀, min (a₁+ N, 3)], N ∈ [0,3]；

Or, if the complexity of the current coded unit is less than the minimum value of the complexity of the adjacent encoder unit, really The depth bounds of the fixed current coded unit is [max (0, a₀- N), a₁], N ∈ [0,3]；

Or, if the complexity of the current coded unit is between the complexity of the adjacent encoder unit, it is determined that institute The depth bounds for stating current coded unit is [a₀, a₁]。

10. the HEVC intraframe coding systems as claimed in claim 6 based on spatial coherence, it is characterised in that the selection Module specifically for：According to the depth and size of current coded unit, select in the left and top of the current coded unit It is adjacent and with an equal amount of adjacent encoder unit, and by calculating the depth bounds of left and top adjacent encoder unit Union obtain the depth bounds [a0, a1] of the adjacent encoder unit.