CN110365982B - Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding - Google Patents

Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding Download PDF

Info

Publication number
CN110365982B
CN110365982B CN201910698624.8A CN201910698624A CN110365982B CN 110365982 B CN110365982 B CN 110365982B CN 201910698624 A CN201910698624 A CN 201910698624A CN 110365982 B CN110365982 B CN 110365982B
Authority
CN
China
Prior art keywords
mode
coding
mts
cand
rate
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910698624.8A
Other languages
Chinese (zh)
Other versions
CN110365982A (en
Inventor
张昊
符婷
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Priority to CN201910698624.8A priority Critical patent/CN110365982B/en
Publication of CN110365982A publication Critical patent/CN110365982A/en
Application granted granted Critical
Publication of CN110365982B publication Critical patent/CN110365982B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/119—Adaptive subdivision aspects, e.g. subdivision of a picture into rectangular or non-rectangular coding blocks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/146—Data rate or code amount at the encoder output
    • H04N19/147—Data rate or code amount at the encoder output according to rate distortion criteria
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/186—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a colour or a chrominance component
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
    • H04N19/96—Tree coding, e.g. quad-tree coding

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

The embodiment of the invention provides a multi-transform selection accelerating method for intra-frame coding in multipurpose coding, which is divided into two stages: the first stage is 'quick skip', the second stage is 'quick termination', wherein the first stage aims at prejudging whether MTS is needed or not, if the prejudged CU does not need to carry out MTS, the rate distortion optimization of MTS is quickly skipped, otherwise, the second stage is entered, and in the second stage, the rate distortion calculation of candidate transformation of some intra-frame prediction modes is considered to be terminated in advance when the MTS is executed, so that the execution time of the transformation process is reduced as much as possible, the problem of long encoding time of a VVC standard encoder VTM in the prior art is solved, the encoding time is shortened, and the applicability is expanded.

Description

Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding
Technical Field
The invention belongs to the technical field of video coding, and particularly relates to a multi-transform selection acceleration method for intra-frame coding in multipurpose coding.
Background
Although High Efficiency Video Coding (HEVC) is currently the main global Video Coding technology, it still cannot meet the Video application and Video requirements that are being developed. Therefore, in 10 months of 2015, a Joint Video encoding Team (Joint Video encoding Team, abbreviated as jviet) was established by a Moving Picture Experts Group (MPEG) and a Video Coding Experts Group (VCEG), and a new generation Video encoding technology is researched to exceed the compression efficiency of HEVC and adapt to the rapid development of the current Video service.
At the 10 th jfet meeting in 4 months in 2018, the jfet defines the first draft of a new generation of Video Coding technology, and names the new Video standard as multifunctional Video Coding (VVC), and issues a corresponding encoder test model VTM 1.0. Different from a quadtree Coding structure of HEVC, VTM1.0 adopts quadtree and a nested multi-type tree structure (QTMT), so that the partitioning of Coding units (Coding units, abbreviated as CUs) is more flexible. From 7 months to 10 months in 2018, jvt successively released second and third drafts of VVC, and the corresponding test models were also updated to VTM2.0 and VTM 3.0.
With the development of video coding technology, many new coding tools, such as a partition structure for separating luminance and chrominance and a Multiple Transform Selection (MTS), are adopted by the VVC and integrated into a VTM (video Test model), which greatly improve the coding efficiency of the VTM, but also greatly increase the time complexity of coding, especially intra-frame coding. The intra coding time complexity of VTM is about several times that of HEVC. The high complexity not only hinders the further development and research process of the VVC, but also is not beneficial to popularization and application in the future.
Disclosure of Invention
The embodiment of the invention solves the problem of long encoding time of a VVC standard encoder VTM in the prior art by providing a multi-transform selection acceleration method for intra-frame encoding in multipurpose encoding, shortens the encoding time and expands the applicability.
A multi-transform selection acceleration method for intra-frame coding in multipurpose coding comprises the following steps:
s1: after the current CU is subjected to DCT-II transformation, judging whether the current CU is the last sub-CU, and if so, performing step S2;
s2: determining whether the CU and the sub-CUs thereof described in step S1 satisfy the formula:
Figure BDA0002150107360000021
if not, go to step S3; if so, go to step S5, where RDpAnd RDi(i is more than or equal to 1 and less than or equal to N) respectively represents the rate distortion cost of the parent CU and the ith child CU, N is the number of the child CUs, and alpha is an adjustable factor and is used for correcting errors;
s3: counting the frequency of each MTS candidate in the adjacent block of the CU, and sorting the MTS candidate list L from high to low according to the frequency of each MTS candidate in the adjacent block of the CU;
s4: the jth candidate transformation of the MTS candidate list L is set as candjJ has an initial value of 0, sets the prediction mode list to S, has a length of N, N is an integer greater than 0, modeiFor the ith prediction mode in S, with an initial value of 0 for i, will be isSkip (mode)i) As whether to skip modeiRate-distortion computation flag, isSkip (mode)i) Are all 0, RDminAs rate-distortion cost minimum, RD, for each mode at each candidate transformationminIs a sufficiently large number, e.g. 1.79769e +308, RD (mode)i,candj) Is a modeiIn candjRate-distortion cost obtained by coding calculation under transform, if j>3, performing step S5;
s5: the MTS procedure is ended.
In step S2, RD ispAnd RDi(1 ≦ i ≦ N) representing the rate-distortion cost of the parent CU and its ith child CU, respectively, N being the number of child CUs, and RD herepRefers to the minimum rate-distortion cost of the parent CU at the current stage. For example, when calculating the rate-distortion cost of a child CU of a parent CU under vertical binary partitioning, RDpMinimum rate-distortion cost for parent CU no partition and horizontal binary partition. According to different division modes, the value of N can be 2 (binary division), 3 (three-fork division) and 4 (four-fork division). RDdctThe rate-distortion cost of the DCT-II transform for the current CU. Alpha is an adjustable factor for correcting errors and takes a real number between 0 and 1, and particularly, when the value of alpha is 0.76, better balance between coding quality and coding time saving can be achieved. Rate-distortion cost RD of DCT-II transform of the last sub-CU when performing MTS of that sub-CUdctRate-distortion cost RD of other sub-CUsi(1≤i<N) and rate distortion cost RD of parent CUpAre all known.
In step S3, the CU neighbor block positions are reordered with respect to the MTS candidate list L as shown in fig. 1, and the process proceeds to step S4.
Wherein steps S2 to S4 are the key steps of the present invention, and the decision of skipping MTS in advance and terminating MTS in advance in the steps plays a key role in reducing the encoding time.
The initialization operation in step S4 is performed only once, and it is not necessary to perform initialization when going to S4, as long as j >3 is determined.
Preferably, in step S1, if no, step S3 is performed.
Preferably, in step S2, the value of the adjustable factor α is a real number between 0 and 1.
Preferably, in step S4, if no, pair candjThe following steps are carried out: and judging whether j is more than or equal to 1 and the adjacent blocks of the CU all select DCT-II transformation, if j is more than or equal to 1 and the adjacent blocks of the CU all select DCT-II transformation, performing step S5.
Further preferably, if j ≧ 1 is not satisfied or all adjacent blocks of the CU are not selected for DCT-II transform, it is determined whether i is greater than or equal to N, if i ≧ N, i is set to 0 and j is added by 1, proceeding to step S4, where proceeding to S4, and the initialization step in S4 is not necessary as long as it is determined whether j is greater than 3 and the subsequent operations.
Still more preferably, if i < N, the following steps are performed for pattern i in S:
judging whether j is greater than or equal to 1 and is (mode)i) If so, then i is incremented by 1 and the process goes to step S4.
Still further preferably, if not, for modeiBy candjCoding is carried out to obtain the corresponding rate distortion cost RD (mode)i,candj) And judging RD (mode)i,candj) Whether or not it is less than RDminIf yes, then RD will be usedminIs updated to RD (mode)i,candj) And i is incremented by 1 and then goes to step S4.
Even more preferably, if RD (mode)i,candj)≥RDminThen isSkip (mode)i) Is updated to 1, i is incremented by 1, and the process proceeds to step S4.
In the intra-frame coding module of VTM, flexible QTMT partition structure and newly introduced coding tools such as MTS and the like are the main reasons for the increase of coding complexity.
Like HEVC and h.264/AVC, VVC also employs a block-based hybrid coding framework. Based on the framework, the VVC coding technology has corresponding module optimization for various redundancies. Fig. 2 shows a typical VVC video encoding flow. As shown in fig. 2, an input picture is first divided into square tiles of equal size, which are called Coding Tree Units (CTUs), which are root nodes of a quad-Tree and nested multi-type Tree division structure. The CTU further divides the partition structure according to the quadtree and the nested multi-type tree into Coding Units (CU), where a CU is a basic Unit for an encoder to perform subsequent processing on a video signal. A CU first performs intra prediction or inter prediction based on its intra-frame and inter-frame properties. If the prediction is intra-frame prediction, the pixel prediction value of the current CU is obtained by mainly utilizing the reference pixels adjacent in space through linear interpolation, and if the prediction is inter-frame prediction, the pixel prediction value of the current CU is obtained by utilizing the reference pixels adjacent in time (the previous frame or the previous frames) through displacement compensation. And then subtracting the original value from the predicted value of the CU to obtain a residual error, and transforming the residual error to further reduce the spatial correlation of the errors of adjacent pixel points and obtain a corresponding residual error coefficient. After the residual coefficient is quantized, entropy coding is carried out by combining information such as a coding mode and related coding parameters, so that a compressed code stream is obtained. On the other hand, the quantized residual coefficient is subjected to inverse quantization and inverse transformation, then the residual and the predicted value are added to obtain a reconstructed pixel, and the reconstructed image is filtered to generate a reference frame and stored in a decoded image buffer to be used as a reference pixel in the following CU intra-frame prediction or inter-frame prediction.
The VVC video coding standard adopts a more flexible structure in the image division technology, namely a structure based on a quadtree and a nested multi-type tree, and supports binary division and trigeminal division in different directions besides the quadtree division. The VVC has five partition modes, namely a quad partition, a vertical binary partition (SPLIT _ BT _ VER), a horizontal binary partition (SPLIT _ BT _ HOR), a vertical trigeminal partition (SPLIT _ TT _ VER), and a horizontal trigeminal partition (SPLIT _ TT _ HOR), wherein the quad partition is called a quad tree structure, and the vertical binary partition, the horizontal binary partition, the vertical trigeminal partition, and the horizontal trigeminal partition are called a multi-type tree structure. Each division mode in the multi-type tree structure is shown in fig. 3-6, fig. 3 is a schematic diagram of vertical binary division, fig. 4 is a schematic diagram of horizontal binary division, fig. 5 is a schematic diagram of vertical trigeminal division, and fig. 6 is a schematic diagram of horizontal trigeminal division. In fig. 3 and 4, the binary partition of CU is a symmetric partition, and in fig. 5 and 6, the trifurcate partition is a partition with a side length of 1:2: 1. Under the multi-type tree division structure, the VVC can better adapt to video sequences with larger and larger resolution, richer and richer contents and more complex textures.
When an image enters the VVC encoder, it is first divided into equal CTUs, which is the largest coding unit and is typically 128 × 128 pixels in size. The QTMT division technology carries out four-fork division by taking the CTU as a root node, and leaf nodes of the four-fork division can be further divided in a recursion mode by a multi-type tree structure until the leaf nodes are divided to a set minimum value. Each node in this structure is a CU. Except special conditions, each division of the nodes of the multi-type tree can traverse four division modes in the multi-type tree structure, and the nodes of the quadtree can traverse the four division modes in the multi-type tree structure and can also try the quadtree division modes. In the process, the optimal partitioning mode and the optimal partitioning depth need to be selected through rate distortion optimization. Taking an intra-frame CU with the size of 64 multiplied by 64 as an example, sequentially performing intra-frame prediction, four-fork division, horizontal two-fork division, vertical two-fork division, horizontal three-fork division and vertical three-fork division on the CU, respectively calculating intra-frame prediction and rate distortion costs under the five divisions, and selecting a mode with the minimum rate distortion cost as a final coding mode.
In HEVC, other coding blocks employ only DCT-II transform, except that luma blocks of 4 × 4 size within a frame may use DST-VII transform and DCT-II transform. In VVC, MTS is adopted to improve the efficiency of residual transformation besides DCT-II transformation. MTS introduces two new transformation kernels: DST-VII and DCT-VIII, both of which can be used for the vertical or horizontal direction of the residual block. That is, the MTS has four candidate transformations, and table 1 lists the candidate transformation numbers of the MTS and their corresponding transformation kernels.
TABLE 1 MTS candidate transformation sequence numbers and corresponding transformation kernels
Figure BDA0002150107360000051
If the MTS is enabled in the Sequence Parameter Set (SPS), the encoder firstly transforms the residual block by DCT-II to calculate the rate-distortion cost RD-DCTII, and then sequentially checks four candidate transformations of the MTS to respectively obtain the rate-distortion costs of 4 transformations. The transform with the least rate-distortion cost is the optimal transform. Fig. 7 shows the residual transform process for VVC intra coding. As shown in FIG. 7, the VVC intra-CU transform process goes through two stages and uses the CU level MTS flag (MTS flag) to indicate whether the current CU enables MTS. And in the first stage, the MTS flag is 0, the CU is transformed by DCT-II, and in the second stage, the MTS flag is 1, and the CU is sequentially transformed by MTS candidate transformation. The encoder transforms the transform with the lowest chosen rate-distortion cost to the optimal transform.
The introduction of MTS improves the coding efficiency of VVC, but it is time consuming to search for the best transform from these candidate transforms for each intra prediction mode of the CU. Therefore, VVC also employs two fast algorithms to reduce the complexity of the residual transformation process, which are represented by the dashed lines in fig. 7. As shown, the core idea of the first fast algorithm is: in the first stage (MTS flag is 0), when the rate distortion cost of a CU in a certain intra-frame prediction mode exceeds a certain threshold value, the rate distortion optimization of the second stage is skipped for the mode, and the next mode is directly coded. Since the rate-distortion cost of one prediction mode transformed with DCT-II may be very large when the rate-distortion cost of the mode transformed with MTS is much larger than that of the other mode, and thus the mode may not be the optimal mode, the MTS rate-distortion cost calculation of the mode may be skipped. The core idea of the second fast algorithm is: in the second phase (MTS flag is 1), rate distortion optimization of the remaining MTS candidate transforms is skipped when the Coded Block Flag (CBF) of the CU under a certain MTS candidate transform is 0. This is because when the transform coefficients of a CU are all 0, the encoder sets its CBF to 0, which means that the CU can most effectively reduce the spatial correlation of the adjacent pixel errors under the transform, and therefore, it is not necessary to perform the rate-distortion cost calculation of other transforms. Furthermore, VVC provides that MTS can only be enabled if a CU satisfies the following conditions: (1) the current coding block is a brightness block; (2) the length and width of the current coding block are all less than or equal to 32; (3) the CBF of the coding block transformed with DCT-II is equal to 1. In the prior art, although the VVC effectively reduces the coding complexity caused by the MTS by using the above simplified scheme, the MTS still has a large promotion space.
The invention aims to provide a multi-transform selection accelerating method based on a multipurpose coding intra-frame coding mode, aiming at the defect of long coding time of a VVC standard coder VTM and the defects of the prior art, so as to shorten the coding time and improve the practical applicability of the method.
And when the sum of the rate-distortion costs of the sub-CUs in a certain division mode is larger than or equal to the current minimum rate-distortion cost of the parent CU, the parent CU does not select the division mode. The MTS rate-distortion cost calculation of the last sub-CU in this partition mode is therefore unnecessary. If this can be predicted, the MTS calculation of the last sub-CU can be skipped in advance, saving coding time.
In addition, the rate distortion calculation order of the MTS candidate transformation in the VTM is fixed (from 0 to 3) according to the candidate index, and the rate distortion calculation order of the MTS candidate transformation is dynamically allocated by fully utilizing the spatial correlation of the CU on transformation selection. As shown in fig. 1, blocks a through E are neighboring blocks of the current CU, whose MTS candidate transform indices are all available at the time of encoding the current CU due to the "zigzag" coding order of the VTM (if blocks a through E select the MTS transform). The order of the MTS candidates of the current CU is decided according to the frequency with which each MTS candidate is selected in the neighboring blocks. Due to the spatial correlation of the CU on the transform selection, the higher the frequency with which an MTS candidate is selected by neighboring blocks, the higher the probability that the current CU selects the candidate transform. Therefore, when deciding the rate-distortion calculation order of the MTS candidates of the current CU, placing candidates with high selection probability in the front is beneficial to improve the accuracy of the MTS fast termination algorithm.
According to the spatial correlation of the CU on the transformation selection, in addition to dynamically allocating the rate-distortion calculation order of the MTS candidate transformation, whether the CU selects the MTS transformation can be predicted. For example, when the neighboring blocks a-E in fig. 1 all select the DCT-II transform, illustrating that the probability of the current CU selecting the MTS is low, the sequential MTS candidates may be considered skipped at this time.
On the basis of sorting of the MTS candidate transformation, the rate distortion cost of each prediction mode under each MTS candidate transformation can be considered to be combined to terminate the prediction mode with lower priority or the MTS candidate transformation in advance.
The embodiment of the invention has the beneficial effects
1. The invention provides a multi-transform selection accelerating method for intra-frame coding in multipurpose coding, which is divided into two stages: the method comprises the steps that a first stage is 'quick skip', a second stage is 'quick termination', wherein the first stage aims at prejudging whether MTS is needed or not, if the prejudging CU does not need to carry out the MTS, rate-distortion optimization of the MTS is quickly skipped, otherwise, the second stage is entered, and in the second stage, rate-distortion calculation of candidate transformation of some intra-frame prediction modes is considered to be terminated in advance when the MTS is executed, so that the execution time of the transformation process is reduced as much as possible;
2. the method provided by the invention has simple steps and small calculated amount, and can be conveniently put into practical application.
Drawings
Fig. 1 is a schematic diagram of CU neighbor block locations.
Fig. 2 is a typical VCC video encoding flow diagram.
Fig. 3 is a schematic diagram of vertical binary partitioning.
Fig. 4 is a schematic diagram of horizontal binary partitioning.
Fig. 5 is a schematic diagram of vertical trifurcation division.
Fig. 6 is a schematic diagram of horizontal trifurcation division.
Fig. 7 is a flowchart of MTS rate-distortion calculation.
Detailed Description
The embodiment of the invention solves the problem of long encoding time of a VVC standard encoder VTM in the prior art by providing a multi-transform selection acceleration method based on a multi-purpose encoding intra-frame encoding mode, shortens the encoding time and expands the applicability.
In order to better understand the above technical solutions, the above technical solutions will be described in detail with reference to specific embodiments.
Example 1
A multi-transform selection acceleration method for intra-frame coding in multipurpose coding comprises the following steps:
s1: after the current CU is subjected to DCT-II transformation, judging whether the current CU is the last sub-CU, and if so, performing step S2;
s2: determining whether the CU and the sub-CUs thereof described in step S1 satisfy the formula:
Figure BDA0002150107360000071
if not, go to step S3; if so, go to step S5, where RDpAnd RDi(i is more than or equal to 1 and less than or equal to N) respectively represents the rate distortion cost of the parent CU and the ith child CU, N is the number of the child CUs, and alpha is an adjustable factor and is used for correcting errors;
s3: counting the frequency of each MTS candidate in the adjacent block of the CU, and sorting the MTS candidate list L from high to low according to the frequency of each MTS candidate in the adjacent block of the CU;
s4: converting the jth candidate of L in step S3 into candjJ has an initial value of 0, sets the prediction mode list to S, has a length of N, N is an integer greater than 0, modeiFor the ith prediction mode in S, with an initial value of 0 for i, will be isSkip (mode)i) As whether to skip modeiRate-distortion computation flag, isSkip (mode)i) Are all 0, RDminAs rate-distortion cost minimum, RD, for each mode at each candidate transformationminIs a sufficiently large number, e.g. 1.79769e +308, RD (mode)i,candj) Is a modeiIn candjRate-distortion cost obtained by coding calculation under transform, if j>3, performing step S5;
s5: the MTS procedure is ended.
In step S2, it should be noted that RD herepRefers to the minimum rate-distortion cost of the parent CU at the current stage. RDpAnd RDi(1. ltoreq. i. ltoreq.N) represents the parent CU and the parent CUAnd the rate distortion cost of the ith sub-CU is N, wherein N is the number of the sub-CUs. Note that, here, RDpRefers to the minimum rate-distortion cost of the parent CU at the current stage. For example, when calculating the rate-distortion cost of a child CU of a parent CU under vertical binary partitioning, RDpMinimum rate-distortion cost for parent CU no partition and horizontal binary partition. According to different division modes, the value of N can be 2 (binary division), 3 (three-fork division) and 4 (four-fork division). RDdctThe rate-distortion cost of the DCT-II transform for the current CU. Alpha is an adjustable factor for correcting errors and takes a real number between 0 and 1, and particularly, when the value of alpha is 0.76, better balance between coding quality and coding time saving can be achieved. Rate-distortion cost RD of DCT-II transform of the last sub-CU when performing MTS of that sub-CUdctRate-distortion cost RD of other sub-CUsi(1≤i<N) and rate distortion cost RD of parent CUpAre all known.
In step S3, the CU neighbor block position is reordered on the MTS candidate list L as shown in fig. 1, and the process proceeds to step four.
Wherein steps S2 to S4 are the key steps of the present invention, and the decision of skipping MTS in advance and terminating MTS in advance in the steps plays a key role in reducing the encoding time.
The initialization operation in step S4 is performed only once, and it is not necessary to perform initialization when going to S4, as long as j >3 is determined.
In step S1, if no, step S3 is performed.
In step S2, the adjustable factor α takes a real number between 0 and 1.
In step S4, if not, the pair candjThe following steps are carried out: and judging whether j is more than or equal to 1 and the adjacent blocks of the CU all select DCT-II transformation, if j is more than or equal to 1 and the adjacent blocks of the CU all select DCT-II transformation, performing step S5.
If j is not equal to or larger than 1 or the adjacent blocks of the CU are not all DCT-II transformed, judging whether i is larger than or equal to N, if i is equal to or larger than N, setting i to 0 and going to step S4. If i < N, performing the following steps on the mode i in S:
judging whether j is greater than or equal to 1 and is isSkip (mo)dei) If so, then i is incremented by 1 and the process goes to step S4. If not, then for modeiBy candjCoding is carried out to obtain the corresponding rate distortion cost RD (mode)i,candj) And judging RD (mode)i,candj) Whether or not it is less than RDminIf yes, then RD will be usedminIs updated to RD (mode)i,candj) And i is incremented by 1 and then goes to step S4.
If RD (mode)i,candj)≥RDminThen isSkip (mode)i) Is updated to 1, i is incremented by 1, and the process proceeds to step S4.
Example 2
This example was implemented based on the VVC official reference platform VTM3.0 and experiments were performed under the general test conditions of JEVT. On the setting of an encoder, the default setting in All-Intra configuration is used, the video sequence used for testing is 22 sequences of six types recommended by JFET, the B-E type test video sequence is the same as the standard test video sequence of HEVC, the input bit depth is 8 bits, the resolution is different from 1920 x 1080 to 416 x 240, the A1-A2 type test video sequence is VVC new ultra high definition test video sequence, and the input bit depth is 10 bits. The encoding performance is mainly evaluated by two indexes of BDBR (Bjotegaard DeltaBit rate) and TS, and the encoding performance of the algorithm is evaluated by taking the original VTM3.0 encoder as a reference (as shown in Table 2, wherein Y, U, V is Y, U, V BDBR respectively). The BDBR represents the code rate difference of two coding methods under the same objective quality, and is obtained by respectively coding and calculating the code rate and the PSNR of the same video under four QP values (22, 27, 32 and 37). The BDBR can comprehensively reflect the code rate and the quality of the video, and the larger the value of the BDBR is, the higher the code rate of the proposed fast algorithm is compared with the original encoder, and the worse the compression performance of the algorithm is. The TS is used to measure the reduction degree of the fast algorithm to the encoding time based on the original encoder, and is calculated as follows:
Figure BDA0002150107360000091
wherein, TpTo be fastTotal encoding time, T, after embedding of fast algorithm into VTM3.0oIs the total encoding time of the original encoder VTM 3.0.
TABLE 2 results of the experiment
Figure BDA0002150107360000092
Figure BDA0002150107360000101
Overall, the BDBR of the UV component is reduced by 0.04% and 0.05% on average, and is approximately none although its RD performance is improved, i.e., the MTS fast algorithm proposed herein has no effect on the chrominance component. This is because VVC adopts a structure of luminance-chrominance separation division, and MTS is used only for luminance. The RD performance of different sequences of chrominance components varies greatly from individual to individual and appears random. For example, some sequences (e.g., CatRobot1, markeplace) have a reduced UV component BDBR, some sequences (e.g., ParkRunning3, ritualdarce) have an increased RD performance, and even for the same sequence (e.g., FoodMarket4, dayightroad 2, etc.), the RD performance of the UV component is opposite. This is because, in the structure of luma and chroma division, when the chroma components are coded after luma coding, the entropy coded context state is changed by the MTS fast algorithm, and the influence of the change on the bit rate is random, so the RD performance of chroma is not consistent from an example. As can be seen from table 2, the algorithm exhibits a high degree of consistency in the impact of BDBR and coding time for different types of sequences, and embodies the robustness and universality of the algorithm.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications, additions and substitutions for the described embodiments may be made by those skilled in the art without departing from the spirit or scope of the invention as defined in the accompanying claims.

Claims (8)

1. A method for accelerating multiple transform selection for intra coding in multipurpose coding, comprising the steps of:
s1: after the current CU is subjected to DCT-II transformation, judging whether the current CU is the last sub-CU, and if so, performing step S2;
s2: determining whether the CU and the sub-CUs thereof described in step S1 satisfy the formula:
Figure FDA0003312260100000011
if not, go to step S3; if so, go to step S5, where RDpAnd RDi(i is more than or equal to 1 and less than or equal to N) respectively represents the rate distortion cost of the parent CU and the ith child CU, N is the number of the child CUs, alpha is an adjustable factor and is used for correcting errors, RDdctRate-distortion cost of the DCT-II transform for the current CU;
s3: counting the frequency of each MTS candidate in the adjacent block of the CU, and sorting the MTS candidate list L from high to low according to the frequency of each MTS candidate in the adjacent block of the CU;
s4: the jth candidate transformation of the MTS candidate list L is set as candjJ has an initial value of 0, sets the prediction mode list to S, has a length of N, N is an integer greater than 0, modeiFor the ith prediction mode in S, with an initial value of 0 for i, will be isSkip (mode)i) As whether to skip modeiRate-distortion computation flag, isSkip (mode)i) Are all 0, RDminAs the minimum value of the rate-distortion cost, RD (mode), for each mode under each candidate transformationi,candj) Is a modeiIn candjRate-distortion cost obtained by coding calculation under transform, if j>3, performing step S5;
s5: the MTS procedure is ended.
2. The method of claim 1, wherein in step S1, if not, step S3 is performed.
3. The method of claim 1, wherein in step S2, the adjustable factor α is a real number between 0 and 1.
4. The method as claimed in claim 1, wherein in step S4, if not, the method is applied to candjThe following steps are carried out: and judging whether j is more than or equal to 1 and the adjacent blocks of the CU all select DCT-II transformation, if j is more than or equal to 1 and the adjacent blocks of the CU all select DCT-II transformation, performing step S5.
5. The method as claimed in claim 4, wherein if j ≧ 1 is unsatisfied or all adjacent CU blocks have not selected DCT-II transform, it is determined whether i is greater than or equal to N, if i ≧ N, i is set to 0 and j is added to 1, and the method proceeds to step S4, where i indicates the ith prediction mode in S described in step S4.
6. The method of claim 5, wherein if i < N, then performing the following steps for mode i in S:
judging whether j is greater than or equal to 1 and is (mode)i) If so, then i is incremented by 1 and the process goes to step S4.
7. The method of claim 6, wherein the multi-transform selection acceleration for intra coding in multipurpose coding,
if not, then for modeiBy candjCoding is carried out to obtain the corresponding rate distortion cost RD (mode)i,candj) And judging RD (mode)i,candj) Whether or not it is less than RDminIf yes, then RD will be usedminIs updated to RD (mode)i,candj) And i is incremented by 1 and then goes to step S4.
8. The method of claim 7, wherein the mode is RD (mode)i,candj)≥RDminThen isSkip (mode)i) Is updated to 1, and after 1 is added to each of i and j, the process proceeds to step S4.
CN201910698624.8A 2019-07-31 2019-07-31 Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding Active CN110365982B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910698624.8A CN110365982B (en) 2019-07-31 2019-07-31 Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910698624.8A CN110365982B (en) 2019-07-31 2019-07-31 Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding

Publications (2)

Publication Number Publication Date
CN110365982A CN110365982A (en) 2019-10-22
CN110365982B true CN110365982B (en) 2022-01-04

Family

ID=68221620

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910698624.8A Active CN110365982B (en) 2019-07-31 2019-07-31 Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding

Country Status (1)

Country Link
CN (1) CN110365982B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021096172A1 (en) * 2019-11-11 2021-05-20 엘지전자 주식회사 Image coding method based on transform, and device therefor
CN111147867B (en) * 2019-12-18 2022-10-18 重庆邮电大学 Multifunctional video coding CU partition rapid decision-making method and storage medium
CN111263158B (en) * 2020-02-28 2021-11-16 中南大学 Multi-transformation-core rapid processing method based on spatial correlation
CN111355955B (en) * 2020-03-06 2022-07-29 中南大学 Multi-transformation core quick skipping method based on pre-selection layer
BR112022014908A2 (en) * 2020-03-27 2022-10-04 Atins Inc VIDEO DECODING METHODS AND VIDEO DECODING DEVICE
CN112689145B (en) * 2020-12-07 2024-02-13 中南大学 Method for selecting quick intra mode of VVC intra coding

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9277212B2 (en) * 2012-07-09 2016-03-01 Qualcomm Incorporated Intra mode extensions for difference domain intra prediction
CN104320658B (en) * 2014-10-20 2017-12-26 南京邮电大学 A kind of HEVC fast encoding methods
US10904526B2 (en) * 2016-03-28 2021-01-26 Kt Corporation Method and apparatus for processing video signal
CN106131547B (en) * 2016-07-12 2018-07-03 北京大学深圳研究生院 The high-speed decision method of intra prediction mode in Video coding
US10972733B2 (en) * 2016-07-15 2021-04-06 Qualcomm Incorporated Look-up table for enhanced multiple transform
US10721489B2 (en) * 2016-09-06 2020-07-21 Qualcomm Incorporated Geometry-based priority for the construction of candidate lists
CN107071418B (en) * 2017-05-05 2020-03-17 上海应用技术大学 HEVC intra-frame coding unit rapid partitioning method based on decision tree
CN109688414B (en) * 2018-12-19 2022-11-11 同济大学 VVC intra-frame coding unit candidate prediction mode reduction and block division early termination method

Also Published As

Publication number Publication date
CN110365982A (en) 2019-10-22

Similar Documents

Publication Publication Date Title
CN110365982B (en) Multi-transformation selection accelerating method for intra-frame coding in multipurpose coding
US11677940B2 (en) Method and device for encoding/decoding image, and recording medium having stored bitstream
US20230336766A1 (en) Image encoding/decoding method and device, and recording medium in which bitstream is stored
US11818340B2 (en) Image encoding/decoding method and device, and recording medium in which bitstream is stored
KR101581100B1 (en) Method for managing a reference picture list, and apparatus using same
EP3817381A1 (en) Image encoding/decoding method and apparatus for throughput enhancement, and recording medium storing bitstream
CN110446036B (en) Coding unit rapid partitioning method based on intra-frame coding in multipurpose coding
KR20190038371A (en) Method and apparatus for encoding/decoding image and recording medium for storing bitstream
CN112369022A (en) Image encoding/decoding method and apparatus, and recording medium storing bit stream
US10652570B2 (en) Moving image encoding device, moving image encoding method, and recording medium for recording moving image encoding program
US20220038682A1 (en) Image encoding/decoding method and device, and recording medium in which bitstream is stored
WO2022117089A1 (en) Prediction method, encoder, decoder and storage medium
WO2009157581A1 (en) Image processing device and image processing method
CN111586405B (en) Prediction mode rapid selection method based on ALF filtering in multifunctional video coding
US20220182661A1 (en) Image encoding/decoding method and apparatus, and recording medium storing bitstream
US20230209051A1 (en) Filtering method and apparatus, and device
US20230132843A1 (en) Image encoding/decoding method and apparatus, and recording medium storing bitstream
KR20220019731A (en) A video encoding/decoding method and apparatus
WO2022116113A1 (en) Intra-frame prediction method and device, decoder, and encoder
CN113841399A (en) Image encoding/decoding method and apparatus
TW202406345A (en) Image processing device and image processing method
CN111355955B (en) Multi-transformation core quick skipping method based on pre-selection layer
KR20160106348A (en) Video Coding Method and Apparatus thereof
WO2022188114A1 (en) Intra-frame prediction method, encoder, decoder, and storage medium
WO2024007450A1 (en) Candidate-list construction method, video coding method, apparatus and system, and video decoding method, apparatus and system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant