CN113014925B

CN113014925B - H.266/VVC inter-frame coding CU fast dividing method and storage medium

Info

Publication number: CN113014925B
Application number: CN202110112537.7A
Authority: CN
Inventors: 李强; 褚嘉璐; 陈昌川; 张栋琪; 余东航; 聂俊
Original assignee: Chongqing University of Post and Telecommunications
Current assignee: Chongqing University of Post and Telecommunications
Priority date: 2021-01-27
Filing date: 2021-01-27
Publication date: 2022-08-23
Anticipated expiration: 2041-01-27
Also published as: CN113014925A

Abstract

The invention relates to a method for quickly dividing H.266/VVC inter-frame coding CU and a storage medium, belonging to the field of video coding, wherein the method comprises the following steps: s1, judging whether the adjacent MVs of the current CU exist and are consistent, if so, entering the next step, and if not, entering S4; s2, adding the current CU coordinates and the MV to obtain the corresponding position coordinates of the reference frame; s3, judging whether the current CU is a square or not, if so, selecting a partial partition mode by using a 9MV model, and if so, selecting a partial partition mode by using a transverse 6MV or longitudinal 6MV model; s4, the CU partition mode initialization is ended. The method can effectively reduce the operation complexity of the H.266/VVC in the CU dividing process, shorten the coding time of the video coder, and can be applied to scenes such as a mobile terminal and the like which need lower video coding operation complexity.

Description

H.266/VVC inter-frame coding CU fast dividing method and storage medium

Technical Field

The invention belongs to the field of Video coding, and particularly relates to a quick Coding Unit (CU) division method in an interframe coding mode in a universal Video coding (VVC) (Video coding), which can be applied to scenes such as a mobile terminal and the like needing low Video coding operation complexity.

Background

The universal Video Coding (VVC), a new generation of Video Coding standard jointly established by MPEG of ISO/IEC and VCEG of ITU-T, has been promulgated in 7 months in 2020. The standard technology is applied to various commercial scenes such as 4K videos, 8K videos, live and immersive media and the like. The VVC still uses a technology based on block Coding, and improves Coding Efficiency by introducing many new technologies based on a High Efficiency Video Coding (HEVC) framework of the previous generation. Compared with the HEVC Test Model (HM), the coding performance of the VTM10.0 in the Random Access (RA) mode is improved by about 40%, but the coding time is increased by more than 9 times, which makes the VVC difficult to be put into practical application. Therefore, under the conditions of ensuring that the video quality is basically unchanged and the coding performance loss is small, how to greatly reduce the computation complexity of the VVC becomes an important research direction of current video coding researchers.

In video coding, a video sequence is first divided into several Groups Of Pictures (GOP), one GOP has 16 frames Of Pictures, the 1 st frame is an I frame using intra-frame prediction, and the rest are B frames using inter-frame prediction. When a B frame is coded, each frame image is firstly divided into a plurality of CTUs, and then the next division is carried out based on the CTUs. The CTU size of HEVC is 64 × 64, and is divided into CUs by Quad-Tree (QT), and the CUs are further divided into Prediction Units (PU) and Transform Units (TU). And the CTU size of the VVC becomes 128 × 128 and the partitioning of CU, PU and TU is cancelled, i.e., prediction and transform are directly performed after the CTU is partitioned into CU. Fig. 2 is an example of CTU partitioning, where if the encoder decides to perform QT partitioning on a CTU, the encoder performs QT or MTT partitioning on a block of size 64 × 64, but the QT partitioning is not allowed after MTT partitioning until the length or width of the block is equal to 4.

Fig. 3 is a diagram of the QTMTT structure corresponding to fig. 2, which is a partition structure of the coding tree finally selected by the CTU. In the process of partitioning the CTU, Rate-Distortion Optimization (RDO) of an encoder is a Depth First Search (DFS) process, which traverses the probability of each partition and calculates RD-cost, and then selects the partition with the smallest RD-cost as a final result. Because the VVC adopts a more flexible dividing mode, the branches of the RDO needing to traverse the tree are increased sharply along with the deepening of the dividing layer number, and the node of each branch needs to be calculated, so the encoding computation amount is increased greatly. In the CU partitioning process, if the partition mode can be reduced, the coding complexity can be significantly reduced, and the coding speed can be increased.

Disclosure of Invention

The present invention is directed to solving the above problems of the prior art. A fast H.266/VVC inter-frame coding CU partitioning method and a storage medium are provided for reducing the coding time. The technical scheme of the invention is as follows:

a fast partitioning method for H.266/VVC inter-coded CU (coding Unit), comprising the following steps:

s1, acquiring CU coding unit data and reference frame data which are coded by a current coding frame of the H.266/VVC coder, and acquiring a stack for maintaining a division mode in a first-in last-out mode;

s2, judging whether the adjacent MV motion vectors of the current CU are consistent, if so, entering a step S3, otherwise, entering a step S5 and stacking all the division modes;

s3, obtaining the coordinate of the corresponding position of the reference frame by adding the current CU coordinate and the MV coordinate, if the CU range of the corresponding position does not exceed the image range, entering the step S4, otherwise, entering the step S5 and putting all the division modes into a stack;

s4, judging whether the current CU is a square or not, if so, selecting a partial partition mode by using a 9MV model, and if so, selecting a partial partition mode by using a transverse 6MV or longitudinal 6MV model;

and S5, finishing initialization of the CU partition mode.

Further, in step S1, CU data being encoded by a current encoded frame of the h.266/VVC encoder and reference frame data are obtained, and a stack in a first-in last-out maintenance partition mode is obtained; in the h.266/VVC encoder, the initialization process of a CU partitioning mode is implemented in an initcurevelle function, and in the initialization process, the encoder may traverse all partitioning modes and push executable partitioning modes into a stack. And then in the rate distortion optimization process of the encoder, the encoder calls the division modes one by one from the stack top to the stack bottom, and selects the mode with the lowest rate distortion cost as the final division mode.

Further, the step S2, determining whether the MV of the current CU is consistent, if yes, going to step S3, otherwise, going to step S5 and stacking all the partition modes includes:

judging whether the adjacent MVs of the current CU are consistent or not, wherein the adjacent MVs are taken from the left CU and the upper CU of the current CU, and if the left CU or the upper CU does not exist due to the fact that the current coding CU is located at the leftmost position or the uppermost position of the image, directly entering step S5 and putting all the partition modes into a stack; and judging whether the MVs are consistent, namely comparing whether the horizontal component and the vertical component of the MVs of the left CU and the upper CU of the current CU are equal, if so, entering the step S3, and if not, entering the step S5 and stacking all the division modes.

Further, the step S3 specifically includes the following steps:

s3, obtaining the coordinate of the corresponding position of the reference frame by adding the current CU coordinate and MV as shown in formula (1) and formula (2), wherein Xcol and Ycol are the coordinate of the corresponding position of the reference frame, Xcur and Ycur are the coordinate of the position of the current frame encoding CU, MVx and MVy are the horizontal component and the vertical component of the MV of the left side CU and the upper CU, if the obtained Xcol and Ycol and the range of the length and the width of the current CU do not exceed the length and the width of the image, the step S3 is entered, otherwise, the step S5 is entered and all the division modes are stacked.

X _col ＝X _cur +MV _x (1)

Y _col ＝Y _cur +MV _y (2)

Further, if the square is obtained in step S4, selecting a partial partition mode using a 9MV model specifically includes: and based on the corresponding position coordinates of the reference frame obtained in the step S3, taking a region with the same size as the current coding CU, and determining which model to use according to the shape of the region, if the region is a square, using a 9MV model, wherein the 9MV model divides a square region into 9 squares with the same size, and the center of the square is used as an MV sampling point. The CU division mode of the H.266/VVC has four choices of four-fork division, two-fork division, three-fork division and no division, wherein the four-fork division is QT, the three-fork division is HTT and VTT, the two-fork division is HBT and VBT, and finally NoSplit is not divided, and the total six modes are six. When a 9MV model is used, judging whether the MVs of 1, 2 and 3 points are consistent or not in the first step; if the two are consistent, entering a further step; if not, the MV of 1, 4 and 7 is judged, if consistent, the HTT and HBT modes are pushed, if not, the QT, VTT and HTT modes which are easy to be divided into a plurality of blocks are pushed. And step two, judging whether the MVs of 4, 5 and 6 are consistent, if so, entering the next step, and otherwise, adding the VTT mode and the VBT mode into the stack. And step three, judging whether the MVs of 7, 8 and 9 are consistent, if so, entering the next step, and otherwise, adding the VTT and VBT modes into the stack. And step four, if all the MVs are consistent, not adding any mode into the stack, otherwise, only adding the VTT mode into the stack.

Further, if the rectangle in step S4 is used, selecting a partial partition mode by using a horizontal 6MV or vertical 6MV model specifically includes:

when using the horizontal 6MV model, in the first step, if 1 and 2 are inconsistent but 1, 3 and 5 are consistent, adding HTT and HBT into the stack; if 1, 3 and 5 are not consistent, adding a VTT mode and an HTT mode; if 1 and 2 are consistent, the next step is carried out. Secondly, if the 3 and 4 are not consistent, adding VTT and VBT modes into a stack; and if the two are consistent, the next step is carried out. And step three, if the 5 and 6 are not consistent, adding the VTT and the VBT into the stack, and otherwise, entering the next step. And fourthly, if all the MVs are consistent, not adding the partitioning mode, and otherwise, only adding the VTT mode into the stack. When a longitudinal 6MV model is used, in the first step, if 1 and 2 are inconsistent and 1, 3 and 5 are inconsistent, adding VTT and HTT modes; if 1, 3 and 5 are consistent, adding VTT and VBT modes; if 1 and 2 are consistent, the next step is carried out. Step two, if the 3 and the 4 are not consistent, the HTT and the HBT are stacked; if the two are consistent, the next step is carried out. Thirdly, if 5 and 6 are not consistent, the HTT and the HBT are put on a stack; otherwise, the next step is carried out. And fourthly, if all the MVs are consistent, not adding the partitioning mode, otherwise, only adding the HTT into the stack.

Further, the step S5 specifically includes: the h.266/VVC inter-frame coding process originally needs to traverse all the partition modes, the CU initialization process puts the modes that can be added on the stack, and after the advance screening of step S4, only part of the modes are put on the stack, and the initialization process is ended, and finally the encoder selects the partition mode with the lowest rate distortion cost from the modes.

A storage medium having stored therein a computer program which, when read by a processor, performs the method of any of claims 1 to 7.

The invention has the following advantages and beneficial effects:

aiming at the problem of overhigh operation complexity of an H.266/VVC encoder, the invention provides a CU (coordinated unit) partitioning rapid algorithm based on a time-space domain. By utilizing the time domain correlation between frames in a video sequence and the spatial domain correlation of the video frames, extracting the MV at the corresponding position of the reference frame by analyzing and using three models of 9MV, horizontal 6MV and vertical 6MV, and screening division modes needing to be traversed by a CU in advance, the problem of overhigh operation complexity caused by traversing all the division modes is avoided. The invention can obviously reduce the encoding time under the condition that the video encoder carries out encoding in an interframe mode, can ensure that the video quality is basically unchanged, has small loss of the encoding output bit rate, and can be applied to scenes such as a mobile terminal and the like which need low video encoding operation complexity. The invention can be applied to the coding of high definition, ultra-high definition, 4K and 8K videos in an inter-frame coding mode of an H.266/VVC video coder, can obviously reduce the complexity of video coding, accelerate the video coding speed and reduce the realization cost of the video coder under the conditions of ensuring that the video quality is basically unchanged and the loss of the coding output bit rate is small, can be applied to application scenes such as a mobile terminal and the like which need low complexity of video coding operation, and has good application prospect and practical value.

The innovation of the present invention is mainly that in step S4 claimed in claim 5 and claim 6, the 9MV, horizontal 6MV and vertical 6MV models are used to extract the reference frame corresponding position MV, and the partition mode of the CU is screened according to the extracted MV features, so as to reduce the CU partition modes that the encoder needs to traverse, and increase the encoding speed. The steps S2 and S3 required by the method of claim 3 and claim 4 are used as a precondition for using the method of step S4, so as to limit the range of use of the method of step S4, and prevent the method from being used in an improper situation to cause the loss of coding performance. The invention has the advantages that the video coding complexity can be obviously reduced, the loss on the output bit rate and the image quality is small, and the invention is easy to realize. According to the existing documents, few researchers screen the CU partition mode according to the MV distribution of the corresponding position of the reference frame, so that the method has strong innovation.

Drawings

FIG. 1 is a flow chart of a fast partitioning method for H.266/VVC inter-frame coded CU according to the preferred embodiment of the present invention;

FIG. 2 is an example of a CU partition for H.266/VVC;

fig. 3 is a diagram of three motion vector point sampling models used in the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be described in detail and clearly with reference to the accompanying drawings. The described embodiments are only some of the embodiments of the present invention.

The technical scheme for solving the technical problems is as follows:

FIG. 1 is a flow chart of a fast partition method for H.266/VVC inter-frame coding CU, according to the invention.

And S1, acquiring CU data which is coded by the current coding frame of the H.266/VVC coder, and acquiring a stack for maintaining the partition mode in a first-in last-out mode by referring to the frame data.

And S2, judging whether the adjacent MVs of the current CU are consistent, wherein the adjacent MVs are taken from the left CU and the upper CU of the current CU. If the left CU or the upper CU does not exist because the current coding CU is located at the leftmost or uppermost position of the image, the process directly proceeds to step S5 and pushes all the division modes onto the stack. And judging whether the MVs are consistent, namely comparing whether the horizontal component and the vertical component of the MVs of the left CU and the upper CU of the current CU are equal, if so, entering the step S3, and if not, entering the step S5 and stacking all the division modes.

And S3, adding the current CU coordinates and the MV to obtain the corresponding position coordinates of the reference frame, and adding the current CU coordinates and the MV to obtain the corresponding position coordinates of the reference frame according to the methods shown in the formula (1) and the formula (2). Where Xcol and Ycol are the coordinates of the corresponding location of the reference frame, Xcur and Ycur are the location coordinates of the CU being encoded by the current frame, and MVx and MVy are the horizontal and vertical components of the MVs of the left and top CUs. If the obtained Xcol and Ycol and the range plus the length and width of the current CU do not exceed the length and width of the image, the process proceeds to step S3, otherwise, the process proceeds to S5 and all the division modes are stacked.

X _col ＝X _cur +MV _x (1)

Y _col ＝Y _cur +MV _y (2)

S4, judging whether the current CU is a square or not, if so, selecting a partial partition mode by using a 9MV model shown in (a) in FIG. 3; in case of a rectangle, a partial partition mode is selected using a vertical 6MV or horizontal 6MV model shown in (b) and (c) of fig. 3. In step S4, an area having the same size as the current CU to be coded is selected based on the position coordinates corresponding to the reference frame obtained in step S3, and a determination is made as to which model is to be used according to the shape of the area. If the region is a square, a 9MV model is used, and if the region is a rectangle, a horizontal 6MV or vertical 6MV model is used. The 9MV model divides a square area into 9 squares with the same size, the center of each square is used as an MV sampling point, and the transverse 6MV and longitudinal 6MV models divide a rectangle into 6 rectangles with the same size, and the center of each square is used as an MV sampling point. As shown in fig. 2, the CU partition pattern of h.266/VVC has four choices of four-way partition, two-way partition, three-way partition, and no partition. The four forks are divided into QT, the three forks are divided into a horizontal three-fork division HTT and a vertical three-fork division VTT, the two forks are divided into a horizontal two-fork division HBT and a vertical two-fork division VBT, and finally NoSplit is not divided, so that the six modes are total. When a 9MV model is used, judging whether MVs at 1, 2 and 3 points are consistent or not in the first step, and entering the next step if the MVs at 1, 2 and 3 points are consistent; if not, the MV of 1, 4 and 7 are judged. If the HTT mode and the HBT mode are consistent, the HTT mode and the HBT mode are pushed; if not, the QT, VTT, HTT patterns, which are easily broken into blocks, are stacked. And secondly, judging whether the

MVs

4, 5 and 6 are consistent or not. If the two are consistent, the next step is carried out; otherwise, VTT and VBT modes are added to the stack. And thirdly, judging whether the MVs of 7, 8 and 9 are consistent or not. If the two are consistent, the next step is carried out; otherwise add VTT and VBT modes to the stack. Step four, if all the MVs are consistent, no mode is added to the stack; otherwise only VTT mode is added to the stack. When a transverse 6MV model is used, in the first step, if 1 and 2 are inconsistent but 1, 3 and 5 are consistent, an HTT and an HBT are added into a stack; if 1, 3 and 5 are not consistent, adding a VTT mode and an HTT mode; if 1 and 2 are consistent, the next step is carried out. Secondly, if the 3 and 4 are not consistent, adding VTT and VBT modes into a stack; and if the two are consistent, the next step is carried out. And step three, if the 5 and 6 are not consistent, adding the VTT and the VBT into the stack, and otherwise, entering the next step. Fourthly, if all the MVs are consistent, the division mode is not added; otherwise only VTT mode is added to the stack. When a longitudinal 6MV model is used, firstly, if 1 and 2 are inconsistent and 1, 3 and 5 are inconsistent, adding VTT and HTT modes; if 1, 3 and 5 are consistent, adding VTT and VBT modes; if 1 and 2 are consistent, the next step is carried out. And secondly, if the

numbers

3 and 4 are not consistent, the HTT and the HBT are stacked, and if the numbers are consistent, the next step is carried out. And thirdly, if the 5 and 6 are not consistent, the HTT and the HBT are stacked. Otherwise, the next step is carried out. And fourthly, if all the MVs are consistent, not adding the partitioning mode, otherwise, only adding the HTT into the stack.

S5, the initialization of the CU partition mode is ended. The process of h.266/VVC inter-frame coding originally needs to go through all the divided modes, and the initialization process of the CU puts the modes that can be added on the stack. And through the advance filtering of step S4, only partial patterns are pushed and the initialization process is ended. Finally, the encoder selects an optimal mode from the modes.

The systems, apparatuses, modules or units described in the above embodiments may be specifically implemented by a computer chip or an entity, or implemented by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The above examples are to be construed as merely illustrative and not limitative of the remainder of the disclosure. After reading the description of the present invention, the skilled person can make various changes or modifications to the invention, and these equivalent changes and modifications also fall into the scope of the invention defined by the claims.

Claims

1. A fast partitioning method for inter-coded CUs of H.266/VVC is characterized by comprising the following steps:

s3, obtaining the position coordinate corresponding to the reference frame by adding the current CU coordinate and the MV, if the CU range of the corresponding position does not exceed the image range, entering a step S4, otherwise, entering a step S5 and stacking all the division modes;

s4, judging whether the current CU is a square or not, if so, selecting a partial partition mode by using a 9MV model, and if so, selecting a partial partition mode by using a transverse 6MV or longitudinal 6MV model; the 9MV model divides a square area into 9 squares with the same size, the center of each square is used as an MV sampling point, the transverse 6MV model and the longitudinal 6MV model divide a rectangle into 6 rectangles with the same size, and the center of each square is used as an MV sampling point;

s5, finishing the initialization of the CU partition mode;

if the step S4 is square, selecting a partial partition mode using a 9MV model; the method specifically comprises the following steps: taking an area with the same size as the current coding CU based on the corresponding position coordinates of the reference frame obtained in the step S3, and determining which model is used according to the shape of the area; if the area is a square, a 9MV model is used, wherein the 9MV model divides a square area into 9 squares with the same size, and the center of each square is used as an MV sampling point; the CU partition mode of the H.266/VVC has four choices of four-fork partition, two-fork partition, three-fork partition and no partition, wherein the four-fork partition is QT, the three-fork partition is HTT and VTT, the two-fork partition is HBT and VBT, and finally NoSplit is not partitioned; when a 9MV model is used, judging whether the MVs of 1, 2 and 3 points are consistent or not in the first step; if the MVs are consistent, entering the next step, if the MVs are not consistent, judging whether the MVs 1, 4 and 7 are consistent, if so, stacking the HTT mode and the HBT mode, and if not, stacking the QT mode, the VTT mode and the HTT mode which are easily divided into a plurality of blocks; secondly, judging whether the MVs of 4, 5 and 6 are consistent, if so, entering the next step, and if not, adding the VTT and VBT modes into a stack; thirdly, judging whether the MVs of 7, 8 and 9 are consistent, if so, entering the next step, and if not, adding the VTT and VBT modes into the stack; if all the MVs are consistent, no mode is added into the stack, otherwise, only the VTT mode is added into the stack;

if the rectangle is obtained in step S4, selecting a partial partition mode by using a horizontal 6MV or vertical 6MV model, which specifically includes:

when a transverse 6MV model is used, in the first step, if 1 and 2 are inconsistent but 1, 3 and 5 are consistent, adding HTT and HBT into a stack; if 1, 3 and 5 are not consistent, adding a VTT mode and an HTT mode; if the 1 and the 2 are consistent, entering the next step; secondly, if the 3 and 4 are not consistent, adding a VTT mode and a VBT mode into a stack, and if the 3 and 4 are consistent, entering the next step; step three, if the 5 and 6 are not consistent, adding VTT and VBT to enter a stack, otherwise, entering the next step; step four, if all the MVs are consistent, not adding the partitioning mode, or only adding the VTT mode into the stack; when a longitudinal 6MV model is used, firstly, if 1 and 2 are inconsistent and 1, 3 and 5 are inconsistent, adding VTT and HTT modes; if 1, 3 and 5 are consistent, adding VTT and VBT modes; if the 1 and the 2 are consistent, entering the next step; step two, if the numbers 3 and 4 are not consistent, the HTT and the HBT are stacked, and if the numbers are consistent, the next step is carried out; thirdly, if the 5 and 6 are not consistent, the HTT and the HBT are stacked, otherwise, the next step is carried out; and step four, if all the MVs are consistent, not adding the partitioning mode, otherwise, only adding the HTT into the stack.

2. An H.266/VVC inter-frame coded CU fast partition method according to claim 1, wherein said step S1, obtaining the CU data being coded by the current coding frame of the H.266/VVC coder and the reference frame data, obtaining the stack for first-in last-out maintenance partition mode; in an H.266/VVC encoder, the initialization process of a CU partition mode is realized in an initCULevel function, in the initialization process, the encoder traverses all partition modes and pushes the partition modes which can be executed into a stack, then in the rate distortion optimization process of the encoder, the encoder calls the partition modes one by one from the top of the stack to the bottom of the stack, and selects the mode with the lowest rate distortion cost as a final partition mode.

3. The method as claimed in claim 1, wherein the step S2 is implemented by determining whether neighboring MVs of a current CU are consistent, if yes, going to step S3, otherwise going to step S5 and stacking all partition modes, and specifically includes:

judging whether the adjacent MVs of the current CU are consistent, wherein the adjacent MVs are taken from the left CU and the upper CU of the current CU, and if the left CU or the upper CU does not exist because the current coding CU is positioned at the leftmost position or the uppermost position of the image, directly entering step S5 and stacking all the division modes; and judging whether the MVs are consistent, namely comparing whether the horizontal component and the vertical component of the MVs of the left CU and the upper CU of the current CU are equal, if so, entering a step S3, and otherwise, entering a step S5 and pushing all the division modes to the stack.

4. The H.266/VVC inter-frame coded CU fast partition method according to claim 3, wherein said step S3 comprises the steps of:

s3, adding the current CU coordinates and the MV to obtain the coordinates of the corresponding position of the reference frame, and adding the current CU coordinates and the MV to obtain the coordinates of the corresponding position of the reference frame as shown in a formula (1) and a formula (2), wherein Xcol and Ycol are the coordinates of the corresponding position of the reference frame; xcur and Ycur are the position coordinates of the CU being encoded by the current frame; MVx and MVy are the horizontal and vertical components of the MVs of the left and top CUs; if the obtained Xcol and Ycol and the range of the length and width of the current CU do not exceed the length and width of the image, the process proceeds to step S4, otherwise, the process proceeds to S5 and all the partition modes are stacked;

X _col ＝X _cur +MV _x (1)

Y _col ＝Y _cur +MV _y (2) 。

5. the method according to claim 1, wherein the step S5 specifically includes: the h.266/VVC inter-frame coding process originally needs to traverse all the partition modes, the CU initialization process puts the modes that can be added on the stack, and through the advanced screening in step S4, only part of the modes are put on the stack, and the initialization process is ended, and finally the encoder selects one partition mode with the lowest rate-distortion cost from the modes as the final partition mode.

6. A storage medium having a computer program stored therein, wherein the computer program, when read by a processor, performs the method of any of claims 1 to 5.