RU2799709C1 - Method and device for video encoding based on long-term reference frame, computing device and data carrier - Google Patents

Abstract

FIELD: image processing technology.

SUBSTANCE: invention relates to video coding based on a long-term reference frame. The method includes the following steps: setting the long-term reference frame according to the attribute information of the image frame; determining a reference index for an image frame to be encoded based on the normal reference frame and the long-term reference frame; and obtaining a target matching block by performing inter-prediction based on the reference index to encode an image frame to be encoded; wherein adjusting the long-term reference frame according to the attribute information of the image frame includes the step of adjusting the long-term reference frame based on the redundancy rate in the time domain and the redundancy rate in the spatial domain for the image frame.

EFFECT: increasing the efficiency of video compression.

12 cl, 7 dwg, 1 tbl

Description

Link to related application

[0001] This application is a national phase transfer in Russia of International Application No. PCT/CN2021/088586, filed April 21, 2020, according to which priority is claimed in accordance with Chinese Patent Application No. BASED ON LONG-TERM REFERENCE FRAME, COMPUTING DEVICE AND STORAGE MEDIA”, the description of which is hereby incorporated by reference in its entirety.

The field of technology to which the present invention relates

[0002] Embodiments of the present disclosure are related to the field of image processing technologies and, in particular, are related to a method and apparatus for encoding video based on a long-term reference frame, a computing device, and a storage medium.

Background of the Invention

[0003] Inter-prediction coding is an important technique widely used in video coding to improve compression efficiency. According to this method, the optimal matching block is searched for in the reference frame by motion estimation, then the prediction block for the current block is predicted by motion compensation technology, so that a residual block is obtained, and the subsequent processing for video encoding compression is realized as a transformation, quantization and entropy coding of the residual block.

Brief summary of the present invention

[0004] The present disclosure provides a method and apparatus for encoding video based on a long-term reference frame, a computing device, and a storage medium.

[0005] According to some embodiments of the present disclosure, a method is provided for encoding video based on a long-term reference frame. This method includes the following steps: setting a long-term reference frame according to image frame attribute information; determining a reference index for an image frame to be encoded based on the normal reference frame and the long-term reference frame; and obtaining a target matching block by performing inter-prediction based on the reference index to encode an image frame to be encoded; wherein setting the long-term reference frame according to the attribute information of the image frame includes the steps of: setting the long-term reference frame based on the time domain redundancy rate and the spatial domain redundancy rate for the image frame, the time domain redundancy rate indicating the degree of presence of the same encoded macroblock in image frame and its adjacent image frame, and the degree of redundancy in the spatial domain indicates the degree of presence of the same encoded macroblock within the image frame.

[0006] According to some embodiments of the present disclosure, an apparatus is provided for encoding video based on a long-term reference frame. This apparatus includes: a setting module configured to set a long-term reference frame according to image frame attribute information; an index module, configured to determine a reference index for an image frame to be encoded based on a normal reference frame and a long-term reference frame; and an encoding module, configured to obtain a target matching block by performing interprediction based on the reference index to encode an image frame to be encoded; wherein the tuning module is further configured to adjust the long-term reference frame based on the degree of redundancy in the time domain and the degree of redundancy in the spatial domain for the image frame, and the degree of redundancy in the time domain indicates the degree of presence of the same encoded macroblock in the image frame and in its adjacent image frame, and the degree of redundancy in the spatial domain indicates the degree of presence of the same encoded macroblock within the image frame.

[0007] According to some embodiments of the present disclosure, a computing device is provided for encoding video based on a long-term reference frame. This computing device includes one or more processors; and a memory configured to store one or more programs, wherein the one or more programs, when loaded and executed by the one or more processors, causes the one or more processors to execute a method for encoding video based on a long-term reference frame, as determined according to any of the embodiments. of this disclosure.

[0008] According to some embodiments of the present disclosure, a non-volatile computer-readable storage medium is provided. The non-volatile computer-readable storage medium stores one or more computer programs, wherein the one or more computer programs, when downloaded and executed by the processor of the computing device, cause the computing device to execute a method for encoding video based on a long-term reference frame, as defined in accordance with any of the embodiments. of this disclosure.

Brief description of the figures

[0009] FIG. 1 shows a flowchart of a method for encoding video based on a long-term reference frame according to Embodiment 1 of the present disclosure;

[0010] FIG. 2 shows a flowchart of a method for encoding video based on a long-term reference frame according to Embodiment 2 of the present disclosure;

[UN] FIG. 3 is a flowchart of a method for encoding video based on a long-term reference frame according to Embodiment 3 of the present disclosure;

[0012] FIG. 4 shows a simplified scheme for searching for a location of an encoded macroblock in a first reference index according to Embodiment 3 of the present disclosure;

[0013] FIG. 5 shows an example of a prediction mode circuit according to Embodiment 3 of the present disclosure;

[0014] FIG. 6 is a simplified block diagram of an apparatus for encoding video based on a long-term reference frame according to Embodiment 4 of the present disclosure; And

[0015] FIG. 7 is a simplified block diagram of a video encoding computing device according to Embodiment 5 of the present disclosure.

Detailed disclosure of the present invention

[0016] The present disclosure is described in detail below with reference to the accompanying figures in conjunction with embodiments. It should be understood that the embodiments described herein are only intended to explain the present disclosure, and are not intended to limit the present disclosure. In addition, it should be further noted that for ease of description, the accompanying figures show only some, and not all, of the structures relevant to the present disclosure.

[0017] Embodiment 1

[0018] FIG. 1 shows a flowchart of a method for encoding video based on a long-term reference frame according to Embodiment 1 of the present disclosure. This embodiment may be applied in a case where a long-term reference frame is used to perform image encoding. The method may be performed by a device for encoding video based on a long-term reference frame, the device being implemented in hardware and/or software. As shown in FIG. 1, the method proposed according to this embodiment of the present disclosure for encoding video based on a long-term reference frame may include the following steps.

[0019] In step 101, the long-term reference frame is adjusted according to the image frame attribute information.

[0020] An image frame can be any image frame in a video that needs to be encoded. The attribute information may be data describing a frame of an image. The attribute information may be stored in association with an image frame. The attribute information may indicate whether the corresponding image frame is a keyframe or a script switch frame. The attribute information may be created by the parsing module in the prior art, and the creation of the attribute information is not described in detail in this embodiment of the present disclosure. There are two types of reference frames for inter-prediction coding: a short-term reference frame and a long-term reference frame. Compared with the short-term reference frame, the long-term reference frame can provide a longer reference picture, the interval of two long-term reference frames in the time domain is longer, and the compression efficiency of coding using the long-term reference frame is better for video with little background change. The long-term reference frame may be a reference picture frame stored in a decoded picture buffer (DPB), and the DPB may contain a picture that is being encoded and may store a short-term reference frame, a long-term reference frame, a non-reference frame, and the like. However, the video coding scheme known to the inventor only functionally supports the long-term reference frame, which greatly increases the coding complexity and reduces the rate of compression coding.

[0021] According to embodiments of the present disclosure, a long-term reference frame may be adjusted according to attribute information for each image frame from a plurality of image frames in a video. For example, in the case where the degree of redundancy in the spatial domain of an image frame exceeds a threshold, the image frame may be set as a long-term image frame. In the case where the attribute information for the image frame indicates that the image frame is the first image frame in the video, the image frame may be set as a long-term image frame.

[0022] In step 102, based on the normal reference frame and the long-term reference frame, a reference index of the image frame to be encoded is determined.

[0023] The normal reference frame may be a reference frame for interprediction in the prior art, and the normal reference frame may be an image frame adjacent in time sequence to the image frame to be coded, for example, it may be the previous frame or the next frame for the frame to be coded Images. The image frame to be encoded may be an image frame to be encoded into a video, and one image frame to be encoded may be encoded each time in the video encoding process. The reference index may be information about the relative location at the frame level of the matched block obtained from the reference frame adjusted based on the image frame to be encoded, and may include a normal reference frame index number at the frame level and/or a long-term reference frame index number at the frame level. frame.

[0024] According to this embodiment of the present disclosure, a reference frame that is most suitable for inter-prediction for an image frame to be coded may be searched for in a normal reference frame and in a long-term reference frame, and the reference frame index number at the frame level may be stored in as a reference index. It may be appreciated that the reference frame that is most suitable for inter-prediction for the image frame to be coded may be the image frame with the lowest encoding overhead in the normal reference frame and/or the long-term reference frame, or may be the image frame closest to to the image frame to be encoded in time sequence in a normal reference frame and/or in a long-term reference frame. The reference index may include at least one of the frame level normal reference frame index numbers and/or the frame level long term reference frame index number.

[0025] In step 103, a target matching block is obtained by performing inter-prediction based on a reference index to perform encoding of an image frame to be coded.

[0026] The target matching block may be an encoded block that is the same or similar to an encoded block in an image frame to be encoded, having a minimum encoding overhead, and that was found by performing inter-prediction based on the reference index.

[0027] An image frame to be encoded may be segmented into at least one encoded macroblock, and for at least one encoded macroblock, a target matching block similar to the encoded macroblock may be found based on the reference index. The image frame to be encoded may be compressed and encoded based on the relevant information about the target matching block. It can be understood that step 103 and step 102 may be performed simultaneously. For example, each time one reference index of an image frame to be encoded is determined, a target matching block corresponding to the image frame to be encoded can be obtained based on the reference frame. Alternatively, step 103 may be performed immediately after complete execution of step 102. For example, immediately after all reference indices of an image frame to be encoded have been determined, a target matching block is searched based on a plurality of reference indices.

[0028] According to this embodiment of the present disclosure, an image frame is set as a long-term reference frame according to attribute information, a reference index of an image frame to be encoded is determined based on the normal reference frame and the long-term reference frame; and obtaining a target matching block for the image frame to be coded is performed by performing inter-prediction based on the reference index to perform encoding of the image frame to be coded. The long-term reference frame is pre-adjusted according to the attribute information of the image frame, thereby achieving fine tuning of the long-term reference frame. The reference index is determined based on the long-term reference frame, and the target matching block is obtained, thereby reducing the computational overhead in the encoding process, reducing the complexity of encoding based on the idea of ensuring compression efficiency in image encoding, reducing the duration of video encoding, and achieving high-efficiency video image encoding .

[0029] Based on the above embodiment of the present disclosure, setting the long-term reference frame according to the image frame attribute information includes: extracting, for each of the image frames from the plurality of image frames, a frame type in the attribute information corresponding to each of the image frames; and setting an image frame corresponding to the frame type as a long-term reference frame in the case where the frame type is a keyframe and/or a script switching frame.

[0030] The frame type may be information describing the usage or content of an image frame. In the case where the frame type is used to describe the use of an image frame, the frame type may include an I key frame, a P non-key frame, a B non-key frame, and the like. In the case where the frame type is used to describe the content, the frame type may include a script switching frame, a script frame, and the like. The keyframe may indicate that the image frame is used as a reference frame in the encoding process, and the script switch frame may indicate that the content of the image frame may include script switch content. For example, in the case where the content of an image frame is an image of a person going outside from a room, the image frame may be a script switching frame.

[0031] Attribute information of each image frame can be obtained from a plurality of image frames in a video. In the case where the frame type in the attribute information for an image frame is a key frame and/or a script switching frame, the image frame may be set as a long-term image frame. For example, a frame image may be stored in a DPB and an identifier for a long-term reference frame may be configured.

[0032] Based on the above embodiment of the present disclosure, setting a long-term reference frame according to image frame attribute information includes: setting a long-term reference frame based on a redundancy rate in the time domain and a redundancy rate in the spatial domain for an image frame.

[0033] According to this embodiment of the present disclosure, the degree of redundancy in the time domain may indicate the degree of presence of the same encoded macroblock in an image frame and in an adjacent image frame, and a large number of the same encoded macroblocks in an image frame and in an adjacent image frame indicates to a higher corresponding degree of redundancy in the time domain. The degree of redundancy in the spatial domain may indicate the degree of presence of the same encoded macroblock within an image frame, and a large number of the same encoded macroblocks within an image frame indicates a higher corresponding degree of redundancy in the spatial domain. Since the long-term reference frame needs to be used as a reference frame for a long time, and the degree of redundancy in the spatial domain of the long-term reference frame is relatively high, an image frame whose degree of redundancy in the spatial domain is larger than the degree of redundancy in the time domain can be selected. and set as a long-term reference frame, or the ratio of the redundancy degree in the time domain to the degree of redundancy in the spatial domain is determined, and an image frame whose ratio is less than a threshold is set as the long-term reference frame.

[0034] Embodiment 2

[0035] FIG. 2 shows a flowchart of a method for encoding video based on a long-term reference frame according to Embodiment 2 of the present disclosure. This embodiment of the present disclosure has been described based on the above embodiment. The long-term reference frame is adjusted based on the degree of redundancy in the time domain and the degree of redundancy in the spatial domain of the image frame. As shown in FIG. 2, the proposed method according to this embodiment of the present disclosure for encoding video based on a long-term reference frame includes the following steps.

[0036] At step 201, for each image frame from a plurality of image frames, the cost of inter-frame coding and intra-frame coding, respectively, and the cost of inter-frame coding reflect the degree of redundancy in the time domain, and the cost of intra-frame coding reflect the degree of redundancy in the spatial domain, the degree of redundancy in the time domain is inversely proportional to the cost of inter-frame coding, and the degree of redundancy in the spatial domain is inversely proportional to the cost of intra-frame coding.

[0037] The cost of inter-frame coding may represent the coding complexity that occurs during inter-frame coding of an image frame. In the case where the coding complexity is lower, the cost of inter-frame coding may be lower. The cost of inter-frame coding may be inversely proportional to the degree of redundancy in the time domain, and a higher degree of redundancy in the time domain of an image frame may correspond to a lower cost of inter-frame coding. The cost of intra-frame coding may represent the coding complexity that occurs during intra-frame coding of an image frame. In the case where the coding complexity is lower, the cost of intra-frame coding may be lower. The cost of intra-frame coding may be inversely proportional to the degree of redundancy in the spatial domain, and a higher degree of redundancy in the spatial domain of an image frame may correspond to a lower cost of intra-frame coding.

[0038] According to this embodiment of the present disclosure, the cost of intra-frame coding and the cost of inter-frame coding can be determined by the preliminary analysis module of the coding system. The intra-frame coding cost and the inter-frame coding cost may be stored together with the image frame as attribute information. During video encoding, image frame attribute information may be obtained in advance. The intra-frame coding cost may reflect a degree of redundancy in the spatial domain of an image frame, and the cost of inter-frame coding may reflect the degree of redundancy in the time domain of an image frame.

[0039] In step 202, the current encoding cost ratio for the current image frame is determined based on the values of the intra-frame coding cost and the inter-frame coding cost. The current image frame is one frame from a plurality of image frames in the video to be encoded.

[0040] The current encoding cost ratio may be the ratio of intra-frame coding costs to inter-frame coding costs, and may be used to measure the amount of redundancy in the spatial domain and the amount of redundancy in the time domain of an image frame. Redundancy is inversely proportional to the cost of coding. For example, the smaller the current value of the encoding cost ratio, the higher the degree of spatial redundancy of the image frame relative to the degree of redundancy in the time domain of the image frame, and during inter-frame coding, the image frame can be used as a long-term reference frame.

[0041] According to this embodiment of the present disclosure, the ratio of intra-frame coding cost to inter-frame coding cost of each image frame of a plurality of image frames in a video may be calculated as a current value of the encoding cost ratio of each image frame, and the inter-frame coding cost may reflect the degree redundancy in the time domain of an image frame, the cost of intra-frame coding may reflect the degree of redundancy in the spatial domain of the image frame, and the degree of redundancy is inversely proportional to the cost of encoding.

[0042] In step 203, the first encoding cost ratio of the image frame preceding the current image frame and the second encoding cost ratio of the image frame before the previous image frame are determined, respectively, and the ratio variance is determined for the current encoding cost ratio, the first cost ratio for coding and a second coding cost ratio.

[0043] The previous image frame for the current image frame may be an image frame at a previous moment relative to the current image frame in time series, and the image frame before the previous image frame may be an image frame at a second moment before the current image frame in time series. The first encoding cost ratio may be the ratio of the intra-frame coding cost to the inter-frame coding cost for the previous image frame relative to the current image frame, and the second encoding cost ratio may be the ratio of the intra-frame coding cost to the inter-frame coding cost for the image frame before the previous frame. Images.

[0044] The first two image frames in each image frame in the DPB can be searched, the intra-frame coding cost and the inter-frame coding cost for every two first image frames can be separately obtained, and the first encoding cost ratio of the previous frame can be separately calculated image and the second cost ratio for encoding another image frame for the respective first two image frames. The ratio variance is calculated based on the ratio of the encoding cost of the current image frame, the ratio of the encoding cost of the previous image frame, and the ratio of the encoding cost of another image frame to determine the change in the encoding cost ratio at the position of each image frame.

[0045] In step 204, the current image frame is set as the long-term reference frame in the case where it is determined that the second encoding cost ratio, the first encoding cost ratio, and the current encoding cost ratio decrease sequentially and the ratio variance is less than the threshold value.

[0046] The threshold value may be a value close to zero. In the case where the variance of the ratio is less than the threshold value, we can assume that the variance of the ratio is 0.

[0047] A change in the encoding cost ratio at a certain position of the current image frame may be determined based on the current encoding cost ratio, the first encoding cost ratio, and the second encoding cost ratio. In the case where it is determined that the current encoding cost ratio is less than the first encoding cost ratio and the first encoding cost ratio is less than the second encoding cost ratio, the image frame encoding cost ratios may be progressively reduced. In the case where the ratio variance is smaller than the threshold value, it can be determined that the change in the encoding cost ratio at a certain position in the current picture frame is 0, and it can be determined that the current encoding cost ratio of the current picture frame is the polar point of the cost ratios for encoding multiple adjacent image frames, and the image frame can be used as a reference frame for the adjacent image frame over time. In this case, the current image frame can be set as a long-term reference frame.

[0048] In step 205, the reference index corresponding to the image frame to be encoded is determined based on the normal reference frame and the long-term reference frame, and sequentially according to the first type prediction mode, the second type prediction mode, and the third type prediction mode, the first type prediction mode being at least least includes a MERGE prediction mode and/or a SKIP prediction mode, a second type prediction mode at least includes a 2N*2N prediction mode and/or a VGOSH prediction mode. (DOWD), and the third type prediction mode at least includes a 2N*N prediction mode, an N*2N prediction mode, a 2N*nD prediction mode, a 2N*nU prediction mode, an nR*2N prediction mode, and/or an nL prediction mode *2N.

[0049] There may be multiple prediction modes. For example, the High Efficiency Video Coding (HEVC) standard may contain 10 types of prediction modes, namely MERGE/SKIP/2N*2N/BIDIR/2N*N/N*2N/2N*nD/2N*nU/nR*2N/nL*2N . In the case where the reference index is separately selected in each prediction mode, the computational complexity increases significantly. Therefore, the prediction modes can be classified into a first type prediction mode, a second type prediction mode, and a third type prediction mode, and a reference index for a picture to be encoded by a frame can be sequentially determined in order for the first type prediction mode, the second type prediction mode, and the third prediction mode. type. The reference index for the second type prediction mode can be created by directly using the reference index for the first type prediction mode, and the reference index for the third type prediction mode can be created by directly using the reference index for the first type prediction mode and the reference index for the second type prediction mode . Therefore, the iterative process of creating a reference index can be shortened, and the complexity of creating a reference index is reduced. According to this embodiment of the present disclosure, the prediction modes can be classified into three types, wherein the first type prediction mode at least includes a MERGE prediction mode and/or a SKIP prediction mode, a second type prediction mode at least includes a 2N*2N prediction mode and/or a GOGO prediction mode. (DOVIN), the third type prediction mode at least includes a 2N*N prediction mode, an N*2N prediction mode, a 2N*nD prediction mode, a 2N*nU prediction mode, an nR*2N prediction mode, and/or an nL* prediction mode 2N.

[0050] In step 206, a target matching block is obtained by performing inter-prediction based on a reference index to perform encoding of an image frame to be coded.

[0051] According to this embodiment of the present disclosure, the inter-frame coding cost of each image frame is obtained to reflect the degree of redundancy in the time domain, and the intra-frame coding cost for each image frame is obtained to reflect the degree of redundancy in the spatial domain; the ratio of intra-frame coding costs to inter-frame coding costs of each image frame is used as the current coding cost ratio; the first encoding cost ratio of the previous image frame for each image frame and the second encoding cost ratio of the image frame before the previous image frame are obtained separately, and the ratio variance is determined based on the current encoding cost ratio, the first encoding cost ratio, and the second encoding cost ratio ; and in this case, when it is determined that the current image frame is a polar point based on the current encoding cost ratio, the first encoding cost ratio, the second encoding cost ratio, and the ratio variance, the current image frame is set as a long-term reference frame. Then, a picture frame reference index corresponding to the first type prediction mode, a picture frame reference index corresponding to the second type prediction mode, and a picture frame reference index corresponding to the third type prediction mode are determined based on the normal reference frame and the long-term reference frame, and sequentially according to the prediction mode a first type, a second type prediction mode, and a third type prediction mode. A target matching block is searched based on a reference index to perform encoding of an image frame to be coded. Therefore, the long-term reference frame is finely tuned, the encoding efficiency is improved, the encoding complexity is reduced, and the image compression ratio is improved based on the idea of securing image encoding compression efficiency, thereby reducing bandwidth cost and improving user experience.

[0052] Embodiment 3

[0053] FIG. 3 shows a flowchart of a method for encoding video based on a long-term reference frame according to Embodiment 3 of the present disclosure. This embodiment of the present disclosure has been described based on the above embodiment in which the search range for the inter-prediction mode is reduced so that the encoding complexity is reduced based on the idea of securing compression efficiency. As shown in FIG. 3, the proposed method according to this embodiment of the present disclosure for encoding video based on a long-term reference frame includes the following steps.

[0054] In step 301, the long-term reference frame is adjusted according to the image frame attribute information.

[0055] In step 302, an encoded macroblock of the image frame to be encoded is obtained according to the first type prediction mode, and a first reference index corresponding to the image frame to be encoded in the first type prediction mode is created by combining the target reference index information of the left block, upper left block, upper block , the upper right block and the lower left block of the encoded macroblock, and target reference index information of the reference encoded macroblock whose position corresponds to the encoded macroblock in the time domain, wherein the reference encoded macroblock is located in a normal reference frame and/or in a long-term reference frame.

[0056] An encoded macroblock may be a set of pixels in an image frame to be encoded. For example, 64 pixels may form one encoded macroblock, and the encoded macroblock may be the smallest component to encode an image frame. An encoded reference macroblock may be the smallest component for encoding a long-term reference frame and/or a normal reference frame, and the encoded reference macroblock may be formed by a plurality of pixels. The left block may be an encoded macroblock located on the left side of the encoded macroblock, and the other blocks may be encoded macroblocks located at appropriate locations from the encoded macroblock.

, причем d представляет глубину кодируемого блока, a RefMask_i представляет опорный индекс (i+1)-го блока. Первый блок является левым блоком текущего кодируемого блока 21 в кадре изображения, второй блок является верхним левым блоком текущего кодируемого блока 21 в кадре изображения, третий блок является верхнем блоком текущего кодируемого блока 21 в кадре изображения, четвертый блок является верхним правым блоком текущего кодируемого блока 21 в кадре изображения, пятый блок является нижним левым блоком текущего кодируемого блока 21 в кадре изображения и шестой блок является опорным кодируемым макроблоком, положение которого соответствует кодируемому макроблоку. В стандарте HEVC k=5, то есть всего имеется k+2 кодируемых макроблоков, которые включают в себя левый блок 23, верхний левый блок 24, верхний блок 25, верхний правый блок 26 и нижний левый блок 22 текущего кодируемого макроблока 21 в кадре изображения, и опорный кодируемый макроблок 21 в опорном кадре, положение которого соответствует кодируемому макроблоку. Во время создания опорного индекса опорный кадр может быть передним опорным кадром или задним опорным кадром. Опорный индекс RefMask_i для (i+1)-го блока удовлетворяет условию RefMask_i=1<<RefForwardi+(1<<RefBackward_i)<<16, где RefForward_i представляет передний опорный индекс (i+1)-го блока (номер индекса переднего опорного кадра на уровне кадра), а RefBackward_i представляет задний опорный индекс (i+1)-го блока (номер индекса заднего опорного кадра на уровне кадра). Номер индекса заднего опорного кадра на уровне кадра можно хранить в 16 старших значащих битах первого опорного индекса, а номер индекса переднего опорного кадра на уровне кадра можно хранить в 16 младших значащих битах первого опорного индекса. То есть, в случае, когда опорный кадр, включающий в себя (i+1)-ый блок, является передним опорным кадром, опорный индекс RefMask_i (i+1)-го блока является индексом, получаемым при перемещении 1 влево на RefForward_i, а в случае, когда опорный кадр, включающий в себя (1+1)-ый блок является задним опорным кадром, опорный индекс RefMask_i (i+1)-го блока является индексом, получаемым при перемещении 1 влево на RefBackward_i и затем на 16 битов. Логическая операция ИЛИ выполняется в соответствующих местах опорных индексов шести блоков для получения первого опорного индекса кодируемого макроблока 21, где «+» представляет операцию ИЛИ. В случае, когда опорный кадр, включающий в себя (1+1)-ый блок, является передним опорным кадром, задний опорный индекс (i+1)-го блока равен 0, а в случае когда опорный кадр, включающий в себя (1+1)-ый блок, является задним опорным кадром, передний опорный индекс (i+1)-го блока равен 0.[0057] In the case where the reference index is obtained according to the first type prediction mode, in the JOIN/SKIP prediction mode, motion estimation is not required, motion compensation is performed directly based on the index inter-frame coding information of a neighboring block in the spatial domain, and the block whose position corresponds to macroblock to be coded in the time domain, and the target macroblock with the minimum coding cost, is finally determined as the inter-frame coding information for the macroblock to be coded. During motion compensation, the interframe coding information of the left block, upper left block, upper block, upper right block, and lower left block from the macroblock to be coded can be used in the spatial domain, and the interframe coding information of the reference macroblock to be coded at the same relative position from the macroblock to be coded may be used in the time domain, wherein the time domain reference frame may include a normal reference frame and/or a long-term reference frame. A reference index for a reference encoded macroblock with optimal reference information may be determined as target reference index information, for example, a frame level index number for an encoded macroblock with the lowest encoding cost, or a frame level index number for an encoded macroblock that is closest to the subject. encoding an image frame in time sequence in a normal reference frame and/or in a long-term reference frame. Target reference index information of the left block, top left block, top block, top right block, and bottom left block and target index information of a coded macroblock reference whose position corresponds to the coded macroblock in the time domain may be stored in the first reference index. In FIG. 4 shows a simplified scheme for searching for a location of an encoded macroblock in a first reference index according to Embodiment 3 of the present disclosure. As shown in FIG. 4, a reference frame and a picture frame may be segmented into a plurality of reference macroblocks 11 and a plurality of coded macroblocks 21. The optimal neighbor block reference indexes in the spatial domain for the macroblock to be coded 21 and the reference codeblock whose position corresponds to the coded macroblock in the time domain may be combined together to get the combined pivot index

, where d represents the depth of the encoded block, and RefMask _i represents the reference index of the (i+1)th block. The first block is the left block of the current encoded block 21 in the image frame, the second block is the top left block of the current encoded block 21 in the image frame, the third block is the top block of the current encoded block 21 in the image frame, the fourth block is the top right block of the current encoded block 21 in the picture frame, the fifth block is the lower left block of the current encoded block 21 in the picture frame, and the sixth block is a reference macroblock to be encoded whose position corresponds to the encoded macroblock. In the HEVC standard, k=5, that is, there are a total of k+2 encoded macroblocks, which include left block 23, top left block 24, top block 25, top right block 26, and bottom left block 22 of the current encoded macroblock 21 in the picture frame , and a reference macroblock to be coded 21 in a reference frame whose position corresponds to the macroblock to be coded. During the creation of the reference index, the reference frame may be a leading reference frame or a trailing reference frame. The reference index RefMask _i for the (i+1)th block satisfies the condition RefMask _i =1<<RefForwardi+(1<<RefBackward _i )<<16, where RefForward _i represents the front reference index of the (i+1)th block (number frame-level front reference frame index), and RefBackward _i represents the back reference index of the (i+1)th block (frame-level back reference frame index number). The frame level back reference frame index number may be stored in the 16 most significant bits of the first reference index, and the frame level front reference frame index number may be stored in the 16 least significant bits of the first reference index. That is, in the case where the reference frame including the (i+1)th block is the front reference frame, the reference index RefMask _i of the (i+1)th block is the index obtained by moving 1 to the left by RefForward _i , and in the case where the reference frame including the (1+1)th block is the back reference frame, the reference index RefMask _i of the (i+1)th block is the index obtained by moving 1 to the left by RefBackward _i and then for 16 bits. The logical OR operation is performed at the respective locations of the reference indexes of the six blocks to obtain the first reference index of the encoded macroblock 21, where "+" represents an OR operation. In the case where the reference frame including the (1+1)th block is a front reference frame, the back reference index of the (i+1)th block is 0, and in the case where the reference frame including the (1 +1)th block is a back reference frame, the front reference index of the (i+1)th block is 0.

[0058] In step 303, recursive encoding is performed on an encoded macroblock of an image frame to be encoded, a parent merge reference index is determined based on the target reference index information of the parent encoded block in the encoded macroblock, and a child merge reference index is determined based on the target reference index information of the child encoded block in the encoded macroblock.

[0059] Recursive encoding is a process of repeating performing recursive segmentation of a macroblock to be encoded. That is, during the recursive encoding process, the macroblock to be coded may be segmented further. For example, the size of an encoded macroblock is 64*64, and the encoded macroblock may be segmented into four child encoded macroblocks with a size of 32*32. During recursive encoding, the size of the child encoded block may vary depending on the segmentation depth. A child encoded block may be created by segmenting an encoded macroblock, and a parent encoded block may be an encoded macroblock whose depth is at least one less than the depth of its child encoded block. For example, for a child codebox with a depth of 3, the depth of the parent codebox can be 2, 1, or 0.

[0060] According to this embodiment of the present disclosure, recursive encoding may be performed on an image frame to be encoded, wherein an encoded macroblock of the image frame to be encoded is then segmented further to create child encoded blocks of different depths, and a reference index of each of the child encoded blocks is determined. . For each of the child coding blocks, a reference index of a parent coding block corresponding to each of the coding blocks and a reference index of a child coding block corresponding to each of the child coding blocks can be created. For example, for a child coding block, the reference indices of the left child box, top left child box, top child box, top right child box, and bottom left child box of the current child coding box, and the reference index of the reference child coding box located at the same position in the time domain, as in the current child encoded block, are obtained as reference indices corresponding to the current child encoded block. Thus, the child reference indexes of the union of all child encoded blocks and the parent reference indices of the union of all parent encoded blocks of the current child encoded block are obtained.

где RefMask_j представляет первый опорный индекс родительского кодируемого блока для дочернего кодируемого блока на глубине j. В случае, когда рекурсия может быть последовательно выполнена вниз от дочернего кодируемого блока, после завершения рекурсии может быть получен опорный индекс каждого дочернего кодируемого блока для дочернего кодируемого блока. Дочерний опорный индекс объединения может быть получен посредством сбора опорных индексов множества дочерних кодируемых блоков, а более конкретно, дочерний опорный индекс объединения удовлетворяет условию

, где SplitRef_m представляет опорный индекс для (m+1)-го дочернего кодируемого блока, и максимальное значение m равно 3, что указывает, что кодируемый макроблок сегментирован в четыре дочерних кодируемых блока.[0061] For example, in the case where the depth d of the child coding block is not less than 1, the parent coding block merge reference reference index for the child coding block can be obtained as

where RefMask _j represents the first parent codeblock reference index for the child codeblock at depth j. In the case where the recursion can be sequentially performed down from the child encoded block, after the completion of the recursion, the reference index of each child encoded block for the child encoded block can be obtained. The child union reference index can be obtained by collecting the reference indices of the plurality of child encoded blocks, and more specifically, the child union reference index satisfies the condition

, where SplitRef _m represents a reference index for the (m+1)th child coded block, and the maximum value of m is 3, indicating that the coded macroblock is segmented into four child coded blocks.

[0062] In step 304, the first reference index, the parent union reference index, and the child union reference index are determined as the second reference index corresponding to the image frame to be encoded in the second type prediction mode.

может представлять родительский опорный индекс объединения, SplitRefSum может представлять дочерний опорный индекс объединения, a RefMask_d может представлять первый опорный индекс. Объединенный второй опорный индекс RefMaskSum используется для ограничения поиска в режиме 2N*2N. Опорный индекс в это время адаптивным образом исключает из себя или включает в себя долговременный опорный индекс, вследствие чего эффективность кодирования может быть улучшена без увеличения сложности кодирования.[0063] The second reference index corresponding to the image frame to be encoded in the second type prediction mode can be obtained by combining the first reference index, the parent combining reference index, and the child combining reference index. For example, the collection of the first pivot index, the parent pivot index of the union, and the child pivot index of the union can be defined as the second pivot index, and the second pivot index satisfies the condition

may represent the parent pivot index of the union, SplitRefSum may represent the child pivot index of the union, and RefMask _d may represent the first pivot index. The combined second reference index RefMaskSum is used to limit the search in 2N*2N mode. The reference index at this time adaptively excludes or includes the long-term reference index, whereby the coding efficiency can be improved without increasing the coding complexity.

[0064] In step 305, a third reference index corresponding to an image frame to be encoded is searched, the search is performed based on the mode type of the third type prediction mode in a predetermined index configuration table.

[0065] The predetermined index configuration table may store the reference index information table created in the recursive encoding process, and the predetermined index configuration table may store the prediction mode in association with the reference index obtained in the method corresponding to the prediction mode.

[0066] According to this embodiment of the present disclosure, the search for the corresponding reference index can be performed according to the mode type of the third type prediction mode. For example, in FIG. 5 shows an example of a prediction mode circuit according to Embodiment 3 of the present disclosure. As shown in FIG. 5, the prediction modes may include 2N*N prediction mode, N*2N prediction mode, 2N*nD prediction mode, 2N*nU prediction mode, nR*2N prediction mode, and nL*2N prediction mode.

[0067] Based on the above embodiment of the present disclosure, the correspondence relationship between the mode type and the third reference index in the predetermined index configuration table includes at least one of the following: in the case where the mode type is a 2N*N prediction mode, the upper the third reference index index is determined based on the target reference index information of the first child encoded block and the target reference index information of the second child encoded block in the recursive encoding process, and the lower index of the third reference index is determined based on the target reference index information of the third child encoded block and the target reference information the index of the fourth child encoded block in the process of recursive encoding; in the case where the mode type is N*2N prediction mode, the left index of the third reference index is determined based on the target reference index information of the first child coded block and the target reference index information of the third child coded block in the recursive encoding process, and the right index of the third reference index is determined based on the target reference index information of the second child encoded block and the target reference index information of the fourth child encoded block in the recursive encoding process; in the case where the mode type is a 2N*nD prediction mode, the upper index of the third reference index is equal to the second reference index, and the lower index of the third reference index is determined based on the target reference index information of the third child coded block and the target reference index information of the fourth child coded block in the process of recursive coding; in the case where the mode type is a 2N*nU prediction mode, the upper index of the third reference index is determined based on the target reference index information of the first child coded block and the target reference index information of the second child coded block in the recursive encoding process, and the lower index of the third reference index is equal to the second reference index; in the case where the mode type is the nR*2N prediction mode, the left index of the third reference index is equal to the second reference index, and the right index of the third reference index is determined based on the target reference index information of the second child coded block and the target reference index information of the fourth child coded block in the process of recursive coding; in the case where the mode type is the nL*2N prediction mode, the left index of the third reference index is determined based on information based on the target reference index of the first child coded block and the target reference index information of the third child coded block in the recursive encoding process, and the right index of the third reference index is equal to the second reference index.

[0068] The first child encoded block, the second child encoded block, the third child encoded block, and the fourth child encoded block may be four child encoded blocks created by segmenting an encoded macroblock in a picture frame to be encoded during a recursive encoding process, and four child encoded blocks may respectively be referred to as a first child code block, a second child code block, a third child code block, and a fourth child code block in the order top left, top right, bottom left, and bottom right.

[0069] According to this embodiment of the present disclosure, in the third type prediction mode, a reference index can be obtained based on a position correspondence relationship in a predetermined index configuration table, where the predetermined index configuration table can be shown as the following table:

[0070] UPPER may represent a superscript, LOWER may represent a subscript, LEFT may represent a left index, and RIGHT may represent a right index. SplitRef ₀ , SplitRef ₂ , SplitRef ₃ , SplitRef ₄ , and Ref2N*2N may respectively represent target reference index information of the first child coded block, target reference index information of the second child coded block, target reference index information of the third child coded block, target reference index information of the fourth child encoded block and a second reference index.

[0071] In step 306, a target matching block is searched based on the reference index in the normal reference frame and/or in the long-term reference frame.

[0072] A normal reference frame may be a normal short-term reference frame in the prior art, and a short-term reference frame may be a neighboring frame to an image frame to be encoded. It may be understood that for a picture frame to be encoded, when a target matching block is searched based on a reference index, the target matching block can be searched in both a normal reference frame and a long-term reference frame, wherein the target matching block can be a coded macroblock, similar to the image frame to be encoded in the normal reference frame and/or in the long-term reference frame. For different prediction modes, different reference indexes may be used to find the target matched block. For example, for the first type prediction mode, the first reference index is used to search for the target matching block, for the second type prediction mode, the second reference index is used to search for the target matching block, and for the third type prediction mode, the third reference index is used to find the target matching block.

[0073] In step 307, compression encoding is performed on the image frame to be coded based on the target matching block.

[0074] An image frame to be encoded can be compressed based on information about the relative position of the target matching block and image data in the target matching block, and the image frame to be encoded can be represented in the form of a relative offset of a plurality of target matching blocks, thereby achieving compression encoding. image frame to be encoded.

[0075] According to this embodiment of the present disclosure, a long-term reference frame is adjusted according to attributes of an image frame, an encoded macroblock of an image frame to be encoded, an adjacent encoded macroblock in the spatial domain of such a frame, and an adjacent encoded macroblock in the time domain of such a frame are obtained according to the first type prediction mode, the first reference index is generated based on the target reference index information of the neighboring encoded macroblock in the spatial domain and the target reference index information of the encoded reference macroblock. Recursive encoding is performed on the macroblock to be coded, during the recursive encoding process, a child union reference index and a parent union reference index are created, and a second reference index used for the second type prediction mode is created by combination. The third reference index for the third type prediction mode may be created by querying a predetermined index configuration table using the model type. A target matching block is searched based on the reference index, and encoding of an image frame to be coded is performed based on the target matching block. Therefore, proper search and encoding of the long-term reference frame is achieved, and the target matching block is sequentially searched according to various types of prediction modes, thereby reducing the complexity of obtaining the reference index, narrowing the search range, and improving encoding efficiency and user experience.

[0076] Based on the above disclosure, based on the reference index, searching for a target matching block in a normal reference frame and/or a long-term reference frame includes: in a 2N*2N prediction mode in the case where it is determined that the normal reference frame and /or the long-term key frame is not a target key frame in the Merge prediction mode, narrowing the search range of the target matching block in the normal key frame and/or in the long-term key frame is performed.

[0077] According to this embodiment of the present disclosure, in the case where the search for the target matching block is performed according to the 2N*2N prediction mode, since the current found reference frame is not the optimal reference frame determined in the JOIN prediction mode, the probability that the target matching block can be found in the current key frame is very small, where the current key frame can be a normal key frame or a long-term key frame. To improve search performance, the search range of the target matching block in the current key frame can be narrowed, for example, it can be narrowed to 1/3 of the initial search range and only 1/3 of the key frames in the normal key frame and in the long-term key frame are selected for search.

[0078] Based on the above disclosure, a reference index based search for a target matching block in a normal reference frame and/or a long-term reference frame includes the steps of: searching for a target matching block based on a diamond search algorithm and a reference index in a normal reference frame and/ or in a long-term reference frame in the case where the attribute information of the image frame to be encoded includes B-frame information and non-key frame information.

[0079] In the case where the attribute information about the image frame to be coded includes B-frame information and non-key frame information, the image frame to be coded may be a small B-frame. The role of the image frame to be encoded in the video is not obvious, and a diamond search algorithm can be used to search for a target matching block to improve the search speed for the target matching block.

[0080] Embodiment 4

[0081] In FIG. 6 is a simplified block diagram of an apparatus for encoding video based on a long-term reference frame according to Embodiment 4 of the present disclosure; An apparatus for encoding video based on a long-term reference frame may perform a method for encoding video based on a long-term reference frame proposed in any embodiment of the present disclosure, and include appropriate functional modules for performing the method. The device may be implemented in hardware and/or software and includes a tuning module 401, an index module 402, and an encoding module 403.

[0082] The adjuster 401 is configured to adjust the long-term reference frame according to the image frame attribute information.

[0083] The index module 402 is configured to determine a reference index of an image frame to be encoded based on a normal reference frame and a long-term reference frame.

[0084] The encoding module 403 is configured to obtain a target matching block by performing inter-prediction based on the reference index to perform encoding of an image frame to be encoded.

[0085] According to this embodiment of the present disclosure, the setting module adjusts an image frame as a long-term reference frame according to the attribute information, the index module determines the reference index of the image frame to be encoded based on the long-term reference frame and the normal reference frame, and the encoding module obtains the target matching block on the image frame to be encoded by performing interprediction based on the reference index to encode the image frame to be encoded. The long-term reference frame is pre-adjusted according to the attribute information of the image frame, thereby achieving fine tuning of the long-term reference frame. The reference index is determined based on the long-term reference frame, and a target matching block is obtained, thereby reducing the computational overhead in the encoding process, and reducing the complexity of encoding based on the idea of ensuring compression efficiency in image encoding.

[0086] Embodiment 5

[0087] FIG. 7 is a simplified block diagram of a video encoding computing device according to Embodiment 5 of the present disclosure. As shown in FIG. 7, a computing device includes a processor 50, memory 51, input hardware 52, and output hardware 53. The device may have one or more processors 50. In FIG. 7, one processor 50 is used as an example. The processor 50, memory 51, input hardware 52, and output hardware 53 of the computing device may be connected to each other via a bus or otherwise. In FIG. 7, as an example, the connection is made with a bus.

[0088] As a computer-readable storage medium, the memory 51 may be configured to store one or more software programs, one or more computer-executable programs, and one or more modules, such as modules (i.e., a tuning module 401, an index module 402, and an encoding module 403) of the video encoding apparatus according to Embodiment 4 of the present disclosure. The processor 50, after downloading and executing one or more programs, one or more instructions, or one or more modules stored in the memory 51, performs a plurality of functional applications and data processing of the computing device, that is, performs the above method for encoding video based on the long-term reference frame.

[0089] The memory 51 may include a program storage area and a data storage area. The program storage area may store an operating system and at least one application program required for the function; and the data storage area may store data created according to the use of the terminal and the like. In addition, the memory 51 may include a high-speed random access memory, and may further include non-volatile memory such as at least one magnetic disk memory device, flash memory device, or other non-volatile semiconductor memory device. In some examples, memory 51 may further include memory located remotely from processor 50. Remote memory may be connected to the device via a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile network, and a combination thereof.

[0090] The input hardware 52 may be configured to receive numeric or text input and generate a key input signal related to the user configuring or controlling the device. The output apparatus 53 may include a display device, such as a display screen.

[0091] Embodiment 6

[0092] Embodiment 6 of the present disclosure further provides a non-volatile storage medium that stores one or more computer-executable instructions. The one or more instructions, when loaded and executed by the computer processor, cause the computer processor to execute a method for encoding a video, the method comprising the steps of: setting a long-term reference frame according to image frame attribute information; determining a reference index for an image frame to be encoded based on the normal reference frame and the long-term reference frame; and obtaining a target matching block by performing inter-prediction based on the reference index to encode an image frame to be encoded.

[0093] Of course, the computer-executable instruction stored on the storage medium proposed in this embodiment of the present disclosure is not limited to performing the above method, and may further perform operations related to the video encoding method proposed according to any embodiment of the present disclosure.

[0094] According to the above descriptions of the embodiments, a person skilled in the art can clearly understand that the present disclosure can be implemented using software and the necessary conventional hardware, or, of course, can be implemented using hardware. Based on such an understanding, the technical solutions of the present disclosure, which contribute substantially or partially to the prior art, may be implemented as a software product. The computer software product may be stored on a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, a hard disk drive, or an optical disk drive of a computer, and may contain includes several instructions for directing a computing device (which may be a personal computer, a server, a network device, or the like) to perform methods according to embodiments of the present disclosure.

[0095] It should be noted that, according to the above embodiment of the video encoding apparatus, the plurality of blocks and modules included therein are only divided according to functional logic, but they are not limited to the above separation, other separations can be applied as long as the corresponding functions are implemented. In addition, the names of a plurality of functional blocks are only for the purpose of distinguishing them from each other, and they are not intended to limit the scope of legal protection of the present disclosure. In addition, the term "and/or" when used with a list of elements means any one of these elements or a combination of them, unless the context clearly states otherwise. For example, "X, Y and/or Z" means "X", "Y", "Z", "X and Y", "X and Z", "Y and Z", or "X, Y and Z", or any combination of the above.

Claims (47)

1. A method for encoding video based on a long-term reference frame, comprising the steps: setting the long-term reference frame according to the image frame attribute information; determining a reference index of an image frame to be encoded based on the normal reference frame and the long-term reference frame; And obtaining a target matching block by performing inter-prediction based on the reference index to perform encoding of an image frame to be coded; wherein setting the long-term reference frame according to the image frame attribute information includes the steps of: setting a long-term reference frame based on a time-domain redundancy rate and a spatial-domain redundancy rate for an image frame, the time-domain redundancy indicating the degree of presence of the same encoded macroblock in the image frame and its adjacent image frame, and the redundancy in The spatial domain indicates the extent to which the same encoded macroblock exists within an image frame. 2. The method of claim 1, wherein setting the long-term reference frame according to the image frame attribute information comprises the steps of: extracting, for each image frame from the plurality of image frames, a frame type from attribute information corresponding to the image frame; And setting an image frame corresponding to the frame type as a long-term reference frame in the case where the frame type is at least one frame of the keyframe or the script switching frame. 3. The method of claim 1, wherein setting the long-term reference frame based on the degree of redundancy in the time domain and the degree of redundancy in the spatial domain for the image frame comprises the steps of: determining for each image frame from a plurality of image frames the cost of inter-frame coding and the cost of intra-frame coding, respectively, and the cost of inter-frame coding reflects the degree of redundancy in the time domain, and the cost of intra-frame coding reflects the degree of redundancy in the spatial domain, the degree of redundancy in the time domain is inversely proportional to the cost of inter-frame coding, and the degree of redundancy in the spatial domain is inversely proportional to the cost of intra-frame coding; determining a current encoding cost ratio for the current image frame based on intra-frame coding cost and inter-frame coding cost values; determining a first encoding cost ratio of an image frame previous to the current image frame and a second encoding cost ratio of an image frame before the previous image frame, and determining a ratio variance for the current encoding cost ratio, the first encoding cost ratio, and the second encoding cost ratio. coding; And setting the current image frame as a long-term reference frame when it is determined that the second encoding cost ratio, the first encoding cost ratio, and the current encoding cost ratio decrease sequentially and the ratio variance is less than a threshold value. 4. The method according to any one of paragraphs. 1-3, wherein determining a reference index of an image frame to be encoded based on a normal reference frame and a long-term reference frame includes the steps of: determining, based on the normal reference frame and the long-term reference frame, and sequentially according to the first type prediction mode, the second type prediction mode, and the third type prediction mode, of the reference index corresponding to the image frame to be encoded, wherein the first type prediction mode at least includes a MERGE prediction mode or a SKIP prediction mode, a second type prediction mode at least includes a 2N*2N prediction mode or a BIDP prediction mode, and a third type prediction mode at least includes a 2N*N prediction mode, N*2N prediction mode, 2N*nD prediction mode, 2N*nU prediction mode, nR*2N prediction mode, or nL*2N prediction mode. 5. The method of claim 4, wherein determining, based on the normal reference frame and the long-term reference frame and sequentially according to the first type prediction mode, the second type prediction mode, and the third type prediction mode, the reference index corresponding to the image frame to be encoded, comprises the steps of: obtaining an encoded macroblock of the image frame to be encoded according to the first type prediction mode, and generating a first reference index corresponding to the image frame to be encoded in the first type prediction mode by combining the target reference index information of the left block, the upper left block, the upper block, the upper right block, and the lower left a block of an encoded macroblock and a reference encoded macroblock whose position corresponds to the encoded macroblock in the time domain, the reference encoded macroblock being located in at least one normal reference frame or a long-term reference frame; performing recursive encoding on an encoded macroblock of an image frame to be encoded, determining a parent merge reference index based on the target reference index information of the parent encoded block in the encoded macroblock, and determining a child reference merge index based on the target reference index information of the child encoded block in the encoded macroblock; determining a first reference index, a parent union reference index, and a child union reference index as a second reference index corresponding to the image frame to be encoded in the second type prediction mode; And searching for a third reference index corresponding to an image frame to be encoded in a predetermined index configuration table based on the mode type of the third type prediction mode. 6. The method of claim 5, wherein the mapping relationship between the mode type and the third reference index in the predetermined index configuration table includes at least one of the following steps: in the case where the mode type is a 2N*N prediction mode, the upper index of the third reference index is determined based on the target reference index information of the first child coded block and the target reference index information of the second child coded block in the recursive encoding process, and the lower index of the third reference index is determined based on the target reference index information of the third child encoded block and the target reference index information of the fourth child encoded block in the recursive encoding process; in the case where the mode type is N*2N prediction mode, the left index of the third reference index is determined based on the target reference index information of the first child coded block and the target reference index information of the third child coded block in the recursive encoding process, and the right index of the third reference index is determined based on the target reference index information of the second child encoded block and the target reference index information of the fourth child encoded block in the recursive encoding process; in the case where the mode type is a 2N*nD prediction mode, the upper index of the third reference index is equal to the second reference index, and the lower index of the third reference index is determined based on the target reference index information of the third child coded block and the target reference index information of the fourth child coded block in the process of recursive coding; in the case where the mode type is a 2N*nU prediction mode, the upper index of the third reference index is determined based on the target reference index information of the first child coded block and the target reference index information of the second child coded block in the recursive encoding process, and the lower index of the third reference index is equal to the second reference index; in the case where the mode type is the nR*2N prediction mode, the left index of the third reference index is equal to the second reference index, and the right index of the third reference index is determined based on the target reference index information of the second child coded block and the target reference index information of the fourth child coded block in the process of recursive coding; And in the case where the mode type is the nL*2N prediction mode, the left index of the third reference index is determined based on the target reference index information of the first child coded block and the target reference index information of the third child coded block in the recursive encoding process, and the right index of the third reference index is equal to the second reference index, and the first child code block, the second child code block, the third child code block, and the fourth child code block are respectively the top left child code block, the top right child code block, the bottom left child code block, and the bottom right child code block in the four child code blocks, created by segmenting the encoded macroblock during the recursive encoding process. 7. The method according to claim 1 or 5, wherein obtaining the target matching block is performed by performing inter-prediction based on the reference index to perform encoding of an image frame to be coded, comprises the steps of: searching for a target matching block based on the reference index in at least one frame from the normal reference frame or the long-term reference frame; And executing based on the target mapped compression coding block on the image frame to be coded. 8. The method of claim 7, wherein searching for a target matching block based on the reference index in at least one frame from a normal reference frame or a long-term reference frame comprises the steps of: in the 2N*2N prediction mode, in the case where it is determined that at least one frame from the normal reference frame or the long-term reference frame is not the target reference frame in the MERGE prediction mode, the search range of the target matching block is narrowed down in at least one frame from a normal key frame or a long-term key frame. 9. The method of claim 7, wherein searching for a target matching block based on the reference index in at least one frame from a normal reference frame or a long-term reference frame comprises the steps of: searching based on a diamond search algorithm for a target matching block in at least one frame from a normal reference frame or a long-term reference frame in the case where the attribute information about the image frame to be encoded includes B-frame information and non-key frame information. 10. Device for encoding video, including: a setting module configured to set the long-term reference frame according to the attribute information of the image frame; an index module, configured to determine a reference index of an image frame to be encoded based on a normal reference frame and a long-term reference frame; And an encoding module, configured to obtain a target matching block by performing inter-prediction based on the reference index to perform encoding of an image frame to be encoded; wherein the tuning module is further configured to adjust the long-term reference frame based on the degree of redundancy in the time domain and the degree of redundancy in the spatial domain for the image frame, and the degree of redundancy in the time domain indicates the degree of presence of the same encoded macroblock in the image frame and in its adjacent image frame, and the degree of redundancy in the spatial domain indicates the degree of presence of the same encoded macroblock within the image frame. 11. A computing device for encoding video based on a long-term reference frame, including: one or more processors; And a memory configured to store one or more programs, wherein one or more programs, when downloaded and executed by one or more processors, causes one or more processors to execute a method for encoding video based on a long-term reference frame, as defined in any one of claims. 1-9. 12. A non-volatile computer-readable storage medium that stores one or more computer programs, wherein the one or more computer programs, when loaded and executed by a processor of a computing device, causes the computing device to execute a method for encoding video based on a long-term reference frame, as defined by any of paragraphs. 1-9.

Images