WO2022067952A1 - 变换集合生成方法和装置、机器可读存储介质和机器设备 - Google Patents

变换集合生成方法和装置、机器可读存储介质和机器设备 Download PDF

Info

Publication number
WO2022067952A1
WO2022067952A1 PCT/CN2020/126033 CN2020126033W WO2022067952A1 WO 2022067952 A1 WO2022067952 A1 WO 2022067952A1 CN 2020126033 W CN2020126033 W CN 2020126033W WO 2022067952 A1 WO2022067952 A1 WO 2022067952A1
Authority
WO
WIPO (PCT)
Prior art keywords
prediction residual
residual block
transform
optimal prediction
transform set
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/126033
Other languages
English (en)
French (fr)
Inventor
李娜
张云
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Publication of WO2022067952A1 publication Critical patent/WO2022067952A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/103—Selection of coding mode or of prediction mode
    • H04N19/109—Selection of coding mode or of prediction mode among a plurality of temporal predictive coding modes
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/124—Quantisation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
    • H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/587—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal sub-sampling or interpolation, e.g. decimation or subsequent interpolation of pictures in a video sequence

Definitions

  • the present invention belongs to the technical field of signal processing, and in particular, relates to a method for generating a transform set of a video prediction residual block, an apparatus for generating a transform set of a video prediction residual block, a machine-readable storage medium, and a machine device.
  • the widely used hybrid video coding framework consists of three modules: prediction coding module, transform coding module and entropy coding module.
  • the predictive coding module includes an intra-frame predictive coding module and an inter-frame predictive coding module. In this way, through the intra-frame predictive coding module and the inter-frame predictive coding module, the correlation between spatial adjacent blocks and time-sequential continuous frames is mined, and the space of the video frame is eliminated. and time redundancy.
  • the prediction residual block is compressed into a code stream through transform coding and entropy coding for storage and transmission.
  • KLT Kerhunen-Loeve Transform
  • the transform kernel is obtained by calculating the co-correlation matrix of each prediction residual block and decomposing it.
  • the computational cost and storage cost of KLT limit its wide application and promotion.
  • DCT Discrete Cosine Transform
  • DST Discrete Sine Transform
  • integer DCT and DST become video coding
  • the DCT transform kernel is fixed, and it is difficult to adapt to the prediction residuals of the contents of various video frames.
  • the prediction residual distribution is related to the video frame and the predictive coding mode.
  • the diversity of the content of the video frame and the predictive coding mode adopted by the predictive coding module leads to a diverse distribution of the prediction residual.
  • Multiple transforms are more able to adapt to the prediction residuals than a single transform.
  • diversity reducing the number of coding bits required to predict the residual transform coefficients.
  • the DCT/DST-based multiple transform selection transform scheme is adopted by the latest coding standards.
  • the latest video coding standards combine the multi-transform selection transform scheme and the inseparable two-level transform based on KLT.
  • Data-driven transformation can balance the contradiction between KLT and DCT transformation performance and computational storage cost.
  • Saab Subspace Approximation with Adjusted Bias
  • the Saab transformation can capture Diversity of different prediction residual blocks.
  • prior art video encoders it has not been considered to generate one or more Saab transforms and integrate them with the existing prediction residual block transform set to further improve the transform performance of the existing transform set.
  • the purpose of the present invention is to provide a method for generating a transform set of a video prediction residual block, an apparatus for generating a transform set of a video prediction residual block, a machine-readable storage medium, and a machine device.
  • a method for generating a transform set of video prediction residual blocks includes: generating a plurality of optimal prediction residual blocks according to an original video frame; a set of subspace approximation transforms based on a bias; storing the set of subspace approximation transforms based on adjustment bias in a video encoder and/or a video decoder; integrating the set of subspace approximation transforms based on adjustment bias into the video in the transform set of the encoder and/or the video decoder.
  • the generating a plurality of optimal prediction residual blocks according to the original video frame includes: acquiring the each prediction residual block of the coding unit in each prediction mode, the original video frame includes a plurality of the encoded units; the prediction residual block with the smallest function value among the respective prediction residual blocks corresponding to the encoded unit Determine the optimal prediction residual block corresponding to the coded unit.
  • the obtaining, according to the plurality of optimal prediction residual blocks, a subspace approximation transform set based on adjustment deviations includes: Divide and classify the plurality of optimal prediction residual blocks into a plurality of optimal prediction residual block sets; obtain at least one subspace approximation transform based on adjustment deviation according to the plurality of optimal prediction residual block sets ; dividing and classifying the at least one adjustment bias-based subspace approximation transform into at least one adjustment bias-based subspace approximation transform set.
  • the difference between the optimal prediction residual block in the optimal prediction residual block set and the set clustering average residual block is The distance is the smallest, and the aggregate cluster average residual block is the average of all optimal prediction residual blocks in the optimal prediction residual block set.
  • an apparatus for generating a transform set of video prediction residual blocks which includes: an optimal prediction residual block obtaining module, configured to generate a plurality of optimal prediction residual blocks according to the original video frame; a transform set The obtaining module is used for obtaining the subspace approximate transformation set based on the adjustment deviation according to the plurality of optimal prediction residual blocks; the transformation storage management module is used for storing the subspace approximate transformation set based on the adjustment deviation in the video coding In the encoder and/or the video decoder; a transform set integration module for integrating the adjusted bias-based subspace approximation transform set into the transform set of the video encoder and/or the video decoder.
  • the optimal prediction residual block obtaining module includes: a prediction residual block obtaining unit, configured to obtain the original video Each prediction residual block of the coded unit of the frame in each prediction mode, the original video frame includes a plurality of the coded units; the optimal prediction residual block determination unit is used to determine the corresponding coded units. The prediction residual block with the smallest function value among the prediction residual blocks is determined as the optimal prediction residual block corresponding to the coded unit.
  • the transform set obtaining module includes: a residual block set unit, configured to convert the plurality of optimal prediction residuals The blocks are divided and classified into a plurality of optimal prediction residual block sets; a transform obtaining unit is used to obtain at least one subspace approximation transform based on adjustment deviation according to the plurality of optimal prediction residual block sets; a transformation set unit , for dividing and classifying the at least one adjustment bias-based subspace approximation transform into at least one adjustment bias-based subspace approximation transform set.
  • the difference between the optimal prediction residual block in the optimal prediction residual block set and the aggregate cluster average residual block is The distance is the smallest, and the aggregate cluster average residual block is the average of all optimal prediction residual blocks in the optimal prediction residual block set.
  • a machine-readable storage medium which stores executable instructions, which, when executed, cause the machine to execute the above-mentioned method for generating a transform set of a video prediction residual block.
  • a machine device comprising: at least one processor, and a memory coupled to the at least one processor, the memory storing instructions, when the instructions are executed by the at least one processor At the time, the at least one processor is caused to execute the above-mentioned method for generating a transform set of a video prediction residual block.
  • the beneficial effects of the present invention are as follows: by integrating the acquired subspace approximate transformation set based on adjustment deviation and the existing transformation set into a new transformation set, the transformation performance of the existing transformation set is further improved.
  • FIG. 1 is a flowchart of a method for generating a transform set of a video prediction residual block according to an embodiment of the present invention
  • FIG. 2 is a block diagram of an apparatus for generating a transform set of a video prediction residual block according to an embodiment of the present invention
  • FIG. 3 is a block diagram illustrating a machine apparatus implementing a method for generating a transform set of a video prediction residual block according to an embodiment of the present invention.
  • the term "including” and variations thereof represent open-ended terms meaning “including but not limited to”.
  • the terms “based on”, “depending on” and the like mean “based at least in part on”, “based at least in part on”.
  • the terms “one embodiment” and “an embodiment” mean “at least one embodiment.”
  • the term “another embodiment” means “at least one other embodiment.”
  • the terms “first”, “second”, etc. may refer to different or the same objects. Other definitions, whether explicit or implicit, may be included below. The definition of a term is consistent throughout the specification unless the context clearly dictates otherwise.
  • a method for generating a transform set of video prediction residual blocks includes: generating a plurality of optimal prediction residual blocks according to an original video frame; The residual block obtains a set of subspace approximation transformations based on adjustment deviations; stores the set of subspace approximation transformations based on adjustment deviations in a video encoder and/or a video decoder; converts the subspace approximation transformations based on adjustment deviations
  • the set is integrated into a transform set of the video encoder and/or the video decoder.
  • the method can be applied to different versions of video codecs including prediction modules (intra-frame prediction module and/or inter-frame prediction module), by comparing the Saab transform calculated from different prediction residual block sets with the existing There are transform sets integrated, which are suitable for the transformation of various prediction residual blocks (intra-frame prediction residual blocks and/or inter-frame prediction modules), thereby further improving the compression efficiency of video codecs.
  • the method for generating a transform set of a video prediction residual block may be performed by a machine device, and the machine device may be connected in communication with other devices.
  • Machine equipment may include smartphones, tablet computers, personal computers, cloud service equipment, service equipment, and the like.
  • FIG. 1 is a flowchart of a method for generating a transform set of a video prediction residual block according to an embodiment of the present invention.
  • an original video frame is obtained, and a plurality of optimal prediction residual blocks are generated according to the original video frame.
  • the original video frame includes a plurality of coded units (or coded blocks). That is, a frame of original video may include a plurality of the coded units, and these coded units will be coded in respective prediction modes to generate prediction residual blocks.
  • the prediction residual block may be an intra prediction residual block. In another example, the prediction residual block may also be an inter prediction residual block.
  • each prediction mode is different and the same, so the prediction residual blocks obtained in each prediction mode are also different.
  • a coded unit of the original video frame is coded in a prediction mode to generate a prediction residual block.
  • a coded unit is coded in multiple prediction modes to generate corresponding multiple prediction residual blocks. For example, if a coded unit is coded in 35 prediction modes, corresponding 35 prediction residual blocks are generated.
  • the function values of the multiple prediction residual blocks corresponding to a coded unit are compared, and the prediction residual block with the smallest function value is selected as the The optimal prediction residual block corresponding to the one coded unit.
  • the prediction mode corresponding to the optimal detection prediction residual block is the optimal prediction mode corresponding to the coded unit.
  • the corresponding optimal prediction residual block here is the optimal intra-frame prediction residual block; and when the above-mentioned obtained is the inter-frame prediction residual block, Correspondingly, the optimal prediction residual block here is the optimal inter prediction residual block.
  • each coded unit of the original video frame is encoded to generate a corresponding one optimal prediction residual block, so that a plurality of optimal prediction residual blocks can be generated from the original video frame.
  • a set of subspace approximation transforms ie Saab transforms
  • adjustment bias is obtained from the plurality of optimal prediction residual blocks.
  • the above-obtained multiple optimal prediction residual blocks are divided and classified into multiple optimal prediction residual block sets.
  • the distance between the optimal prediction residual block and the set clustering average residual block is the smallest.
  • the aggregate cluster average residual block here is the average of all optimal prediction residual blocks in each optimal prediction residual block set.
  • multiple optimal prediction residual blocks obtained above may be collected into a total set of optimal prediction residual blocks, and then the total optimal prediction residual block set may be divided into multiple optimal prediction residual blocks A subset of prediction residual blocks.
  • the distance between the optimal prediction residual block and the aggregate cluster average residual block is the smallest.
  • the aggregate cluster average residual block here is the average of all optimal prediction residual blocks in each optimal prediction residual block subset. That is to say, the optimal prediction residual block subset in another example is equivalent to the above-mentioned optimal prediction residual block set.
  • At least one subspace approximation transformation based on adjustment bias is obtained according to the plurality of optimal prediction residual block sets.
  • a corresponding subspace approximation transform based on adjustment bias may be obtained according to an optimal prediction residual block set, so that the number of subspace approximation transforms based on adjustment bias equal to the number of optimal prediction residual block sets is obtained.
  • a corresponding subspace approximation transform based on adjustment bias may be obtained according to at least two optimal prediction residual block sets.
  • the obtained subspace approximation transformation based on the adjustment deviation may be a one-layer transformation, or a cascaded multi-layer transformation.
  • the transformation coefficients of the prediction residual block output by the transformation are one-dimensional transform coefficient vectors.
  • the realization of the one-dimensional transform coefficient vector is beneficial to the design and realization of the quantization coding module, the entropy coding module, etc., and can further reduce the number of compressed bits.
  • the subspace approximation transformation based on the adjustment deviation is a one-layer transformation, then for a prediction residual block of 8*8 resolution, after the subspace approximation transformation based on the adjustment deviation, it becomes 1*64 resolution , so as to realize one-dimensional transformation.
  • the subspace approximation transformation based on the adjustment bias is a cascaded two-layer transformation, where the first layer is a 2*2 transformation layer, and the second layer is a 4*4 transformation layer, then for a prediction of 8*8 resolution
  • the residual block is first divided into 16 sub-blocks of 2*2 resolution, and the 16 sub-blocks of 2*2 resolution become 4 sub-blocks of 4*4 resolution after passing through the 2*2 transformation layer block, and then these four 4*4 resolution sub-blocks become 1*64 resolution sub-blocks after passing through the 4*4 transformation layer, thereby realizing one-dimensional transformation.
  • the at least one adjustment bias-based subspace approximation transform is divided and classified into at least one adjustment bias-based subspace approximation transform set.
  • the set of adjusted bias-based subspace approximation transforms is stored in a video encoder and/or a video decoder.
  • the set of subspace approximation transforms based on adjustment bias is synchronously stored at the video encoder side and/or the video decoder side for backup to manage the set of subspace approximation transforms based on adjustment bias.
  • the adjusted bias-based subspace approximate transform set is integrated into a transform set of the video encoder and/or the video decoder.
  • the adjusted bias-based subspace approximation transform set is integrated with an existing transform set in the video encoder and/or video decoder to form a new transform set that transforms the prediction residual block .
  • the existing transform set here includes at least one existing transform for transforming the prediction residual block.
  • the new transform set obtained in block 107 is used to transform the prediction residual block. Specifically, in one example, given m 1 +m 2 prediction residual blocks, m 1 prediction residual blocks are transformed using the Saab transform in the new transform set obtained in block 107 , while block 107 is used to transform the m 1 prediction residual blocks.
  • the existing transforms in the new transform set obtained in transform the remaining m 2 prediction residual blocks.
  • m 1 prediction residual blocks are transformed using the Saab transform in the new transform set obtained in block 107 and the existing transform, and The remaining m 2 prediction residual blocks are transformed using the existing transforms in the new transform set obtained in block 107 .
  • m 1 prediction residual blocks are transformed using the Saab transform in the new transform set obtained in block 107 and the existing transform, and The remaining m 2 prediction residual blocks are transformed using the Saab transform in the new transform set obtained in block 107 and the existing transform.
  • FIG. 2 is a block diagram of an apparatus for generating a transform set of a video prediction residual block according to an embodiment of the present invention.
  • the apparatus 200 for generating a transform set of a video prediction residual block includes: an optimal prediction residual block obtaining module 202, a transform set obtaining module 204, a transform storage management module 206, and a transform set integration module 208.
  • the optimal prediction residual block obtaining module 202 is configured to obtain the original video frame, and generate a plurality of optimal prediction residual blocks according to the original video frame.
  • the optimal prediction residual block acquisition module 202 may include a prediction residual block acquisition unit and an optimal prediction residual block determination unit.
  • the original video frame includes a plurality of coded units (or coded blocks). That is, one frame of original video may include a plurality of the encoded units.
  • the prediction residual block obtaining unit is used to encode the coded unit in each prediction mode to generate a prediction residual block.
  • the prediction residual block may be an intra prediction residual block.
  • the prediction residual block may also be an inter prediction residual block.
  • the prediction residual block obtaining unit encodes a coded unit of the original video frame in a prediction mode to generate a prediction residual block.
  • the prediction residual block obtaining unit encodes a coded unit in multiple prediction modes to generate corresponding multiple prediction residual blocks. For example, when the prediction residual block obtaining unit encodes one coded unit in 35 prediction modes, corresponding 35 prediction residual blocks are generated.
  • the optimal prediction residual block determination unit is configured to compare the function values of a plurality of prediction residual blocks corresponding to a coded unit, and to select the prediction residual block with the smallest function value as the one to be coded The optimal prediction residual block corresponding to the unit.
  • the prediction mode corresponding to the optimal detection prediction residual block is the optimal prediction mode corresponding to the coded unit.
  • the corresponding optimal prediction residual block here is the optimal intra-frame prediction residual block; and when the above-mentioned obtained is the inter-frame prediction residual block, Correspondingly, the optimal prediction residual block here is the optimal inter prediction residual block.
  • the optimal prediction residual block obtaining module 202 encodes each coded unit of the original video frame to generate a corresponding optimal prediction residual block, so that a plurality of optimal prediction residual blocks can be generated according to the original video frame Optimal prediction residual block.
  • the transform set obtaining module 204 is configured to obtain a set of subspace approximate transforms (ie Saab transforms) based on adjustment deviations according to the plurality of optimal prediction residual blocks.
  • the transform set obtaining module 204 may include a residual block set unit, a transform obtainment unit, and a transform set unit.
  • the residual block set unit is used to divide and classify the obtained multiple optimal prediction residual blocks into multiple optimal prediction residual block sets.
  • the distance between the optimal prediction residual block and the set clustering average residual block is the smallest.
  • the aggregate cluster average residual block here is the average of all optimal prediction residual blocks in each optimal prediction residual block set.
  • the residual block set unit may first collect the plurality of optimal prediction residual blocks obtained above into a total set of optimal prediction residual blocks, and then divide the total set of optimal prediction residual blocks into is a subset of multiple optimal prediction residual blocks.
  • the distance between the optimal prediction residual block and the aggregate cluster average residual block is the smallest.
  • the aggregate cluster average residual block here is the average of all optimal prediction residual blocks in each optimal prediction residual block subset. That is to say, the optimal prediction residual block subset in another example is equivalent to the above-mentioned optimal prediction residual block set.
  • the transform obtaining unit is configured to obtain at least one subspace approximation transform based on adjustment bias according to the plurality of optimal prediction residual block sets.
  • the transform obtaining unit may obtain a corresponding subspace approximation transformation based on adjustment deviation according to an optimal prediction residual block set, so as to obtain an adjustment deviation-based subspace equal to the number of optimal prediction residual block sets Approximate transformation.
  • the transform obtaining unit may also obtain a corresponding subspace approximation transform based on the adjustment deviation according to the at least two optimal prediction residual block sets.
  • the transform set unit is configured to divide and categorize the at least one adjustment bias-based subspace approximation transform into at least one adjustment bias-based subspace approximation transform set.
  • the transform storage management module 206 is configured to store the set of subspace approximation transforms based on the adjustment bias in the video encoder and/or the video decoder.
  • the transformation storage management module 206 synchronously stores the set of subspace approximation transformations based on adjustment bias at the video encoder side and/or the video decoder side, so as to perform the subspace approximation transformation based on adjustment bias. Collections are managed.
  • a transform set integration module 208 is configured to integrate the adjusted bias-based subspace approximation transform set into a transform set of the video encoder and/or the video decoder.
  • the transform set integration module 208 integrates the adjusted bias-based subspace approximation transform set with an existing transform set in the video encoder and/or video decoder to form a new pair of prediction residual blocks
  • the set of transforms to transform includes at least one existing transform for transforming the prediction residual block.
  • the apparatus for generating a transform set of a video prediction residual block may be implemented by hardware, or may be implemented by software or a combination of hardware and software. Taking software implementation as an example, a device in a logical sense is formed by reading the corresponding computer program instructions in the memory into the memory for operation by the processor of the device where it is located. In an embodiment of the present invention, the apparatus for generating a transform set of a video prediction residual block may be implemented by, for example, a machine device (eg, a computer device).
  • a machine device eg, a computer device
  • FIG. 3 is a block diagram illustrating a machine apparatus implementing a method for generating a transform set of a video prediction residual block according to an embodiment of the present invention.
  • a machine device 300 may include at least one processor 310, memory (eg, non-volatile memory) 320, memory 330, and communication interface 340, and at least one processor 310, memory 320, memory 330, and communication interface 340 Connected together via bus 350 .
  • At least one processor 310 executes at least one computer-readable instruction stored or encoded in memory (ie, the above-described elements implemented in software).
  • computer-executable instructions are stored in memory that, when executed, cause at least one processor 310 to perform a process of: generating a plurality of optimal prediction residual blocks from the original video frame; generating a plurality of optimal prediction residual blocks from the plurality of optimal prediction residuals
  • the difference block obtains a set of subspace approximation transformations based on adjustment deviations; stores the set of subspace approximation transformations based on adjustment deviations in a video encoder and/or a video decoder; stores the set of subspace approximation transformations based on adjustment deviations Integrated into a transform set of the video encoder and/or the video decoder.
  • a program product eg, a machine-readable medium
  • a machine-readable medium may have instructions (ie, the above-described elements implemented in software) that, when executed by a machine, cause the machine to perform various operations and functions described in connection with FIG. 1 above in embodiments of the present invention.
  • a system or an apparatus equipped with a readable storage medium may be provided, on which software program codes for realizing the functions of any of the above-described embodiments are stored, and a computer or a computer of the system or apparatus may be provided.
  • the processor reads and executes the instructions stored in the readable storage medium.
  • the program code itself read from the readable medium can implement the functions of any one of the above-described embodiments, and thus the machine-readable code and the readable storage medium storing the machine-readable code constitute the present invention part of the example.
  • Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (eg, CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tape, non- Volatile memory cards and ROMs.
  • the program code may be downloaded from a server computer or the cloud over a communications network.
  • the device structure described in the above embodiments may be a physical structure or a logical structure, that is, some units may be implemented by the same physical entity, or some units may be implemented by multiple physical entities, or may be implemented by multiple physical entities. Some components in separate devices are implemented together.

Landscapes

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

Abstract

一种视频预测残差块的变换集合生成方法、变换集合生成装置、机器可读存储介质以及机器设备。所述变换集合生成方法包括：根据原始视频帧产生多个最优预测残差块（101）；根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合（103）；将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中（105）；将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中（107）。所述方法通过将所获取的基于调整偏差的子空间近似变换集合与已有的变换集合集成一个新的变换集合，从而进一步提高现有的变换集合的变换性能。

Description

变换集合生成方法和装置、机器可读存储介质和机器设备 技术领域
本发明属于信号处理技术领域，具体地讲，涉及一种视频预测残差块的变换集合生成方法、视频预测残差块的变换集合生成装置、机器可读存储介质以及机器设备。
背景技术
目前，广泛应用的混合视频编码框架由预测编码模块、变换编码模块和熵编码模块三个模块构成。其中，预测编码模块包括帧内预测编码模块和帧间预测编码模块，这样，通过帧内预测编码模块和帧间预测编码模块来挖掘空间相邻块和时序连续帧间相关性，消除视频帧空间和时间冗余。预测残差块经过变换编码和熵编码压缩为码流进行存储和传输。
KLT(Karhunen-Loeve Transform)具有最优能量集中性和去相关性，通过计算每个预测残差块协相关矩阵并分解得到变换核。然而，KLT的计算代价和存储代价限制了其广泛应用和推广。当预测残差符合一阶马尔科夫过程时，离散余弦变换(Discrete Cosine Transform，简称DCT)或离散正弦变换(Discrete Sine Transform，简称DST)的变换性能逼近于KLT，整数DCT和DST成为视频编码器中主要变换。然而，DCT变换核固定，难以适应于多种多样视频帧的内容的预测残差。
目前，变换集合的生成和部署主要依赖于大规模编解码实验。预测残差分布与视频帧和预测编码模式相关，视频帧的内容和预测编码模块采用的预测编码模式的多样性导致预测残差分布多样，多个变换相比于单个变换更能够适应预测残差的多样性，减少预测残差变换系数所需编码比特数。考虑多个变换核能够更好适应预测残差块内容的多样性，基于DCT/DST的多变换选择变换方案被最新的编码标准采纳。考虑DCT和KLT可以捕获到多种多样内容纹理模式，最新得视频编码标准将多变换选择变换方案和基于KLT的不可分二级变换进行联合应用。
数据驱动变换能够平衡KLT和DCT变换性能和计算存储代价之间的矛盾。在卷积神经网络广泛应用背景下，研究工作者提出了一种数据驱动、多层级联和不可分的变换，即基于调整偏差的子空间近似(SubspaceApproximation withAdjusted Bias，简称Saab)变换，Saab变换能够捕获不同预测残差块的多样性。然而，在现有技术的视频编码器中，尚未考虑生成一个或多个Saab变换，并将其与已有的预测残差块变换集合进行集成，从而进一步提高现有变换集合的变换性能。
发明内容
为了解决上述现有技术存在的技术问题，本发明的目的在于提供一种视频预测残差块的变换集合生成方法、视频预测残差块的变换集合生成装置、机器可读存储介质以及机器设备。
根据本发明的一方面提供的视频预测残差块的变换集合生成方法，其包括：根据原始视频帧产生多个最优预测残差块；根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合；将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中；将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
进一步地，在根据本发明的一方面提供的视频预测残差块的变换集合生成方法中，所述根据原始视频帧产生多个最优预测残差块，包括：获取所述原始视频帧的被编码单元在各个预测模式下的各个预测残差块，所述原始视频帧包括多个所述被编码单元；将所述被编码单元对应的各个预测残差块中函数值最小的预测残差块确定为所述被编码单元对应的最优预测残差块。
进一步地，在根据本发明的一方面提供的视频预测残差块的变换集合生成方法中，所述根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合，包括：将所述多个最优预测残差块划分并归类到多个最优预测残差块集合中；根据所述多个最优预测残差块集合获取至少一个基于调整偏差的子空间近似变换；将所述至少一个基于调整偏差的子空间近似变换划分并归类到至少一个基于调整偏差的子空间近似变换集合中。
进一步地，在根据本发明的一方面提供的视频预测残差块的变换集合生成 方法中，所述最优预测残差块集合中的最优预测残差块与集合聚类平均残差块的距离最小，所述集合聚类平均残差块是所述最优预测残差块集中所有最优预测残差块的平均。
根据本发明的另一方面提供的视频预测残差块的变换集合生成装置，其包括：最优预测残差块获取模块，用于根据原始视频帧产生多个最优预测残差块；变换集合获取模块，用于根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合；变换存储管理模块，用于将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中；变换集合集成模块，用于将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
进一步地，在根据本发明的一方面提供的视频预测残差块的变换集合生成装置中，所述最优预测残差块获取模块包括：预测残差块获取单元，用于获取所述原始视频帧的被编码单元在各个预测模式下的各个预测残差块，所述原始视频帧包括多个所述被编码单元；最优预测残差块确定单元，用于将所述被编码单元对应的各个预测残差块中函数值最小的预测残差块确定为所述被编码单元对应的最优预测残差块。
进一步地，在根据本发明的一方面提供的视频预测残差块的变换集合生成装置中，所述变换集合获取模块包括：残差块集合单元，用于将所述多个最优预测残差块划分并归类到多个最优预测残差块集合中；变换获取单元，用于根据所述多个最优预测残差块集合获取至少一个基于调整偏差的子空间近似变换；变换集合单元，用于将所述至少一个基于调整偏差的子空间近似变换划分并归类到至少一个基于调整偏差的子空间近似变换集合中。
进一步地，在根据本发明的一方面提供的视频预测残差块的变换集合生成装置中，所述最优预测残差块集合中的最优预测残差块与集合聚类平均残差块的距离最小，所述集合聚类平均残差块是所述最优预测残差块集中所有最优预测残差块的平均。
根据本发明的再一方面提供的机器可读存储介质，其存储有可执行指令，所述指令当被执行时使得所述机器执行上述的视频预测残差块的变换集合生成方法。
根据本发明的又一方面提供的机器设备，其包括：至少一个处理器，以及与所述至少一个处理器耦合的存储器，所述存储器存储指令，当所述指令被所述至少一个处理器执行时，使得所述至少一个处理器执行上述的视频预测残差块的变换集合生成方法。
本发明的有益效果：通过将所获取的基于调整偏差的子空间近似变换集合与已有的变换集合集成一个新的变换集合，从而进一步提高现有的变换集合的变换性能。
附图说明
通过结合附图进行的以下描述，本发明的实施例的上述和其它方面、特点和优点将变得更加清楚，附图中：
图1是根据本发明的实施例的视频预测残差块的变换集合生成方法的流程图；
图2是根据本发明的实施例的视频预测残差块的变换集合生成装置的模块图；
图3是示出了根据本发明的实施例的实现视频预测残差块的变换集合生成方法的机器设备的方框图。
具体实施方式
以下，将参照附图来详细描述本发明的具体实施例。然而，可以以许多不同的形式来实施本发明，并且本发明不应该被解释为限制于这里阐述的具体实施例。相反，提供这些实施例是为了解释本发明的原理及其实际应用，从而使本领域的其他技术人员能够理解本发明的各种实施例和适合于特定预期应用的各种修改。
如本文中使用的，术语“包括”及其变型表示开放的术语，含义是“包括但不限于”。术语“基于”、“根据”等表示“至少部分地基于”、“至少部分地根据”。术语“一个实施例”和“一实施例”表示“至少一个实施例”。术语“另一个实施例”表示“至少一个其他实施例”。术语“第一”、“第二”等可以指代不同的或相同的对象。下面可以包括其他的定义，无论是明确的还是隐含的。 除非上下文中明确地指明，否则一个术语的定义在整个说明书中是一致的。
如背景技术中所述，在现有技术的视频编码器(和/或视频解码器)中，尚未考虑生成一个或多个Saab(Subspace Approximation with Adjusted Bias，基于调整偏差的子空间近似)变换，并将生成的一个或多个Saab变换与已有的预测残差块变换集合进行集成，从而形成一个新的预测残差块变换集合，利用该新的预测残差块变换集合对预测残差块进行变换，可以实现进一步提高现有变换集合的变换性能。为了实现这一目的，根据本发明的实施例提供了视频预测残差块的变换集合生成方法，其包括：根据原始视频帧产生多个最优预测残差块；根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合；将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中；将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
因此，在该方法中，通过将所获取的Saab变换集合与已有的变换集合集成一个新的变换集合，从而进一步提高现有的变换集合的变换性能。此外，进一步地，该方法可以适用于包含预测模块(帧内预测模块和/或帧间预测模块)的不同版本视频编解码器，通过将从不同预测残差块集合计算得到的Saab变换与已有变换集合集成，从而适用于多种多样预测残差块(帧内预测残差块和/或帧间预测模块)的变换，进而进一步提高视频编解码器压缩效率。
以下将结合附图来详细描述根据本发明的实施例的视频预测残差块的变换集合生成方法和变换集合生成装置。
根据本发明的实施例的视频预测残差块的变换集合生成方法可以由机器设备来执行，机器设备可以与其他的设备通信连接。机器设备可以包括智能手机、平板电脑、个人计算机、云服务设备、服务设备等。
图1是根据本发明的实施例的视频预测残差块的变换集合生成方法的流程图。
参照图1，在块101中，获取原始视频帧，并根据原始视频帧产生多个最优预测残差块。
在一个示例中，所述原始视频帧包括多个被编码单元(或称被编码块)。也就是说，一帧原始视频可以包括多个所述被编码单元，这些被编码单元将在各个预测模式下被编码以产生预测残差块。在一个示例中，预测残差块可以是帧内预测残差块。在另一个示例中，预测残差块还可以是帧间预测残差块。
在一个示例中，对一帧原始视频进行编码时所采取的预测模式通常具有多个，例如35个。需要说明的是，各个预测模式各不同相同，从而在各个预测模式下得到的预测残差块也各不相同。在这种情况下，原始视频帧的一个被编码单元在一个预测模式下被编码以产生一个预测残差块。这样，一个被编码单元在多个预测模式下被编码就产生相应的多个预测残差块。例如，一个被编码单元在35个预测模式下被编码就产生相应的35个预测残差块。
在一个示例中，在获取一个被编码单元相应的多个预测残差块之后，对一个被编码单元相应的多个预测残差块的函数值进行比较，选函数值最小的预测残差块作为该一个被编码单元相应的最优预测残差块。需要说明的是，该最优侦测预测残差块所对应的预测模式即为该被编码单元相应的最优预测模式。此外，当上述获取的是帧内预测残差块时，相应的此处的最优预测残差块是最优帧内预测残差块；而当上述获取的是帧间预测残差块时，相应的此处的最优预测残差块是最优帧间预测残差块。
在这种情况下，所述原始视频帧的每个被编码单元被编码以产生相应的一个最优预测残差块，从而可以根据所原始视频帧产生多个最优预测残差块。
在块103中，根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换(即Saab变换)集合。
在一个示例中，将上述得到的多个最优预测残差块划分并归类到多个最优预测残差块集合中。这里，需要进一步强调的是，划分归类后，在每个最优预测残差块集合中，最优预测残差块与集合聚类平均残差块的距离最小。其中，这里的集合聚类平均残差块是每个最优预测残差块集合中所有最优预测残差块的平均。
在另外一个示例中，可以先将上述得到的多个最优预测残差块收集到一个最优预测残差块总集合中，而后将该最优预测残差块总集合划分为多个最优预 测残差块子集合。这里，在每个最优预测残差块子集合中，最优预测残差块与集合聚类平均残差块的距离最小。其中，这里的集合聚类平均残差块是每个最优预测残差块子集合中所有最优预测残差块的平均。也就是说，在另外一个示例中的最优预测残差块子集合等同于上述的最优预测残差块集合。
在得到如上所述的多个最优预测残差块集合之后，在一个示例中，根据所述多个最优预测残差块集合获取至少一个基于调整偏差的子空间近似变换。在一个示例中，可以根据一个最优预测残差块集合获取相应的一个基于调整偏差的子空间近似变换，如此得到与最优预测残差块集合数量相等的基于调整偏差的子空间近似变换。在另一个示例中，可以根据至少两个最优预测残差块集合获取相应的一个基于调整偏差的子空间近似变换。
进一步地，需要说明的是，所得到的基于调整偏差的子空间近似变换可以是一层变换，也可以是级联的多层变换。更进一步地，利用所述基于调整偏差的子空间近似变换对预测残差块进行变换时，变换所输出的预测残差块变换系数是一维变换系数向量。而一维变换系数向量的实现，有益于进行量化编码模块、熵编码编码模块等的设计和实现，能够进一步减少压缩比特数。
例如，基于调整偏差的子空间近似变换是一层变换，那么对于一个8*8分辨率的预测残差块，其通过该基于调整偏差的子空间近似变换之后，变成为1*64分辨率的子块，从而实现一维变换。再例如，基于调整偏差的子空间近似变换是级联的两层变换，其中第一层为2*2变换层，第二层为4*4变换层，那么对于一个8*8分辨率的预测残差块，先将其划分为16个2*2分辨率的子块，16个2*2分辨率的子块通过2*2变换层之后，变成为4个4*4分辨率的子块，而后这4个4*4分辨率的子块通过4*4变换层之后变成为1*64分辨率的子块，从而实现一维变换。
在得到如上所述的至少一个基于调整偏差的子空间近似变换之后，将所述至少一个基于调整偏差的子空间近似变换划分并归类到至少一个基于调整偏差的子空间近似变换集合中。
在块105中，将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中。
在一个示例中，将所述基于调整偏差的子空间近似变换集合在视频编码器端和/或视频解码器端同步存储备份，以对所述基于调整偏差的子空间近似变换集合进行管理。
在块107中，将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
在一个示例中，将所述基于调整偏差的子空间近似变换集合与视频编码器和/或视频解码器中已有的变换集合进行集成，以形成新的对预测残差块进行变换的变换集合。需要说明的是，这里的已有的变换集合中包括至少一个已有的对预测残差块进行变换的变换。
进一步地，块107中得到的新的变换集合被用于对预测残差块进行变换。具体地，在一个示例中，给定m 1+m 2个预测残差块，使用块107中得到的新的变换集合中的Saab变换对m 1个预测残差块进行变换，而使用块107中得到的新的变换集合中的已有的变换对余下的m 2个预测残差块进行变换。
在另一个示例中，给定m 1+m 2个预测残差块，使用块107中得到的新的变换集合中的Saab变换和已有的变换对m 1个预测残差块进行变换，而使用块107中得到的新的变换集合中的已有的变换对余下的m 2个预测残差块进行变换。
在又一个示例中，给定m 1+m 2个预测残差块，使用块107中得到的新的变换集合中的Saab变换和已有的变换对m 1个预测残差块进行变换，而使用块107中得到的新的变换集合中的Saab变换和已有的变换对余下的m 2个预测残差块进行变换。
以下对于根据本发明的实施例的视频预测残差块的变换集合生成装置进行详细描述。图2是根据本发明的实施例的视频预测残差块的变换集合生成装置的模块图。
参照图2，根据本发明的实施例的视频预测残差块的变换集合生成装置200包括：最优预测残差块获取模块202、变换集合获取模块204、变换存储管理模块206以及变换集合集成模块208。
最优预测残差块获取模块202用于获取原始视频帧，并根据原始视频帧产生多个最优预测残差块。在一个示例中，最优预测残差块获取模块202可以包括预测残差块获取单元和最优预测残差块确定单元。
在一个示例中，所述原始视频帧包括多个被编码单元(或称被编码块)。也就是说，一帧原始视频可以包括多个所述被编码单元。预测残差块获取单元用于在各个预测模式下对被编码单元进行编码以产生预测残差块。在一个示例中，预测残差块可以是帧内预测残差块。在另一个示例中，预测残差块还可以是帧间预测残差块。在一个示例中，对一帧原始视频进行编码时所采取的预测模式通常具有多个，例如35个。需要说明的是，各个预测模式各不同相同，从而在各个预测模式下得到的预测残差块也各不相同。在这种情况下，预测残差块获取单元对原始视频帧的一个被编码单元在一个预测模式下编码以产生一个预测残差块。这样，预测残差块获取单元在多个预测模式下对一个被编码单元进行编码就产生相应的多个预测残差块。例如，预测残差块获取单元在35个预测模式下对一个被编码单元进行编码就产生相应的35个预测残差块。
在一个示例中，最优预测残差块确定单元用于对一个被编码单元相应的多个预测残差块的函数值进行比较，并用于选函数值最小的预测残差块作为该一个被编码单元相应的最优预测残差块。需要说明的是，该最优侦测预测残差块所对应的预测模式即为该被编码单元相应的最优预测模式。此外，当上述获取的是帧内预测残差块时，相应的此处的最优预测残差块是最优帧内预测残差块；而当上述获取的是帧间预测残差块时，相应的此处的最优预测残差块是最优帧间预测残差块。
在这种情况下，最优预测残差块获取模块202对所述原始视频帧的每个被编码单元编码以产生相应的一个最优预测残差块，从而可以根据所原始视频帧产生多个最优预测残差块。
变换集合获取模块204用于根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换(即Saab变换)集合。在一个示例中，变换集合获取模块204可以包括残差块集合单元、变换获取单元以及变换集合单元。
在一个示例中，残差块集合单元用于将上述得到的多个最优预测残差块划分并归类到多个最优预测残差块集合中。这里，需要进一步强调的是，划分归 类后，在每个最优预测残差块集合中，最优预测残差块与集合聚类平均残差块的距离最小。其中，这里的集合聚类平均残差块是每个最优预测残差块集合中所有最优预测残差块的平均。
在另外一个示例中，残差块集合单元可以先将上述得到的多个最优预测残差块收集到一个最优预测残差块总集合中，而后将该最优预测残差块总集合划分为多个最优预测残差块子集合。这里，在每个最优预测残差块子集合中，最优预测残差块与集合聚类平均残差块的距离最小。其中，这里的集合聚类平均残差块是每个最优预测残差块子集合中所有最优预测残差块的平均。也就是说，在另外一个示例中的最优预测残差块子集合等同于上述的最优预测残差块集合。
在一个示例中，变换获取单元用于根据所述多个最优预测残差块集合获取至少一个基于调整偏差的子空间近似变换。在一个示例中，变换获取单元可以根据一个最优预测残差块集合获取相应的一个基于调整偏差的子空间近似变换，如此得到与最优预测残差块集合数量相等的基于调整偏差的子空间近似变换。在另一个示例中，变换获取单元也可以根据至少两个最优预测残差块集合获取相应的一个基于调整偏差的子空间近似变换。
在一个示例中，变换集合单元用于将所述至少一个基于调整偏差的子空间近似变换划分并归类到至少一个基于调整偏差的子空间近似变换集合中。
变换存储管理模块206用于将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中。
在一个示例中，变换存储管理模块206将所述基于调整偏差的子空间近似变换集合在视频编码器端和/或视频解码器端同步存储备份，以对所述基于调整偏差的子空间近似变换集合进行管理。
变换集合集成模块208用于将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
在一个示例中，变换集合集成模块208将所述基于调整偏差的子空间近似变换集合与视频编码器和/或视频解码器中已有的变换集合进行集成，以形成新 的对预测残差块进行变换的变换集合。需要说明的是，这里的已有的变换集合中包括至少一个已有的对预测残差块进行变换的变换。
以上参照图1和图2，对根据本发明的实施例的视频预测残差块的变换集合生成方法和视频预测残差块的变换集合生成装置进行了描述。
根据本发明的实施例的视频预测残差块的变换集合生成装置可以采用硬件实现，也可以采用软件或者硬件和软件的组合来实现。以软件实现为例，作为一个逻辑意义上的装置，是通过其所在设备的处理器将存储器中对应的计算机程序指令读取到内存中运行形成的。在本发明的实施例中，使用进行视频预测残差块的变换集合生成的装置例如可以利用机器设备(如计算机设备)来实现。
图3是示出了根据本发明的实施例的实现视频预测残差块的变换集合生成方法的机器设备的方框图。
参照图3，机器设备300可以包括至少一个处理器310、存储器(例如，非易失性存储器)320、内存330和通信接口340，并且至少一个处理器310、存储器320、内存330和通信接口340经由总线350连接在一起。至少一个处理器310执行在存储器中存储或编码的至少一个计算机可读指令(即，上述以软件形式实现的元素)。
在一个示例中，在存储器中存储计算机可执行指令，其当执行时使得至少一个处理器310执行过程：根据原始视频帧产生多个最优预测残差块；根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合；将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中；将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
应该理解，在存储器中存储的计算机可执行指令当执行时使得至少一个处理器310在进行根据本发明的实施例中结合以上图1描述的各种操作和功能。
根据一个实施例，提供了一种例如机器可读介质的程序产品。机器可读介质可以具有指令(即，上述以软件形式实现的元素)，该指令当被机器执行时，使得机器执行本发明的实施例中的结合以上图1描述的各种操作和功能。
具体地，可以提供配有可读存储介质的系统或者装置，在该可读存储介质上存储着实现上述实施例中任一实施例的功能的软件程序代码，且使该系统或者装置的计算机或处理器读出并执行存储在该可读存储介质中的指令。
在这种情况下，从可读介质读取的程序代码本身可实现上述实施例中任何一项实施例的功能，因此机器可读代码和存储机器可读代码的可读存储介质构成了本发明的实施例的一部分。
可读存储介质的实施例包括软盘、硬盘、磁光盘、光盘(如CD-ROM、CD-R、CD-RW、DVD-ROM、DVD-RAM、DVD-RW、DVD-RW)、磁带、非易失性存储卡和ROM。可选择地，可以由通信网络从服务器计算机上或云上下载程序代码。
上述对本发明的特定实施例进行了描述。其它实施例在所附权利要求书的范围内。在一些情况下，在权利要求书中记载的动作或步骤可以按照不同于实施例中的顺序来执行并且仍然可以实现期望的结果。另外，在附图中描绘的过程不一定要求示出的特定顺序或者连续顺序才能实现期望的结果。在某些实施方式中，多任务处理和并行处理也是可以的或者可能是有利的。
上述各流程和各系统结构图中不是所有的步骤和单元都是必须的，可以根据实际的需要忽略某些步骤或单元。各步骤的执行顺序不是固定的，可以根据需要进行确定。上述各实施例中描述的装置结构可以是物理结构，也可以是逻辑结构，即，有些单元可能由同一物理实体实现，或者，有些单元可能分由多个物理实体实现，或者，可以由多个独立设备中的某些部件共同实现。
在整个本说明书中使用的术语“示例性”、“示例”等意味着“用作示例、实例或例示”，并不意味着比其它实施例“优选”或“具有优势”。出于提供对所描述技术的理解的目的，具体实施方式包括具体细节。然而，可以在没有这些具体细节的情况下实施这些技术。在一些实例中，为了避免对所描述的实施例的概念造成难以理解，公知的结构和装置以框图形式示出。
以上结合附图详细描述了本发明的实施例的可选实施方式，但是，本发明的实施例并不限于上述实施方式中的具体细节，在本发明的实施例的技术构思范围内，可以对本发明的实施例的技术方案进行多种简单变型，这些简单变型均属于本发明的实施例的保护范围。
本说明书内容的上述描述被提供来使得本领域任何普通技术人员能够实 现或者使用本说明书内容。对于本领域普通技术人员来说，对本说明书内容进行的各种修改是显而易见的，并且，也可以在不脱离本说明书内容的保护范围的情况下，将本文所定义的一般性原理应用于其它变型。因此，本说明书内容并不限于本文所描述的示例和设计，而是与符合本文公开的原理和新颖性特征的最广范围相一致。

Claims (10)

  1. 一种视频预测残差块的变换集合生成方法，其中，所述变换集合生成方法包括：
    根据原始视频帧产生多个最优预测残差块；
    根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合；
    将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中；
    将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
  2. 根据权利要求1所述的变换集合生成方法，其中，所述根据原始视频帧产生多个最优预测残差块，包括：
    获取所述原始视频帧的被编码单元在各个预测模式下的各个预测残差块，所述原始视频帧包括多个所述被编码单元；
    将所述被编码单元对应的各个预测残差块中函数值最小的预测残差块确定为所述被编码单元对应的最优预测残差块。
  3. 根据权利要求1或2所述的变换集合生成方法，其中，所述根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合，包括：
    将所述多个最优预测残差块划分并归类到多个最优预测残差块集合中；
    根据所述多个最优预测残差块集合获取至少一个基于调整偏差的子空间近似变换；
    将所述至少一个基于调整偏差的子空间近似变换划分并归类到至少一个基于调整偏差的子空间近似变换集合中。
  4. 根据权利要求3所述的变换集合生成方法，其中，所述最优预测残差 块集合中的最优预测残差块与集合聚类平均残差块的距离最小，所述集合聚类平均残差块是所述最优预测残差块集中所有最优预测残差块的平均。
  5. 一种视频预测残差块的变换集合生成装置，其中，所述变换集合生成装置包括：
    最优预测残差块获取模块，用于根据原始视频帧产生多个最优预测残差块；
    变换集合获取模块，用于根据所述多个最优预测残差块获取基于调整偏差的子空间近似变换集合；
    变换存储管理模块，用于将所述基于调整偏差的子空间近似变换集合存储于视频编码器和/或视频解码器中；
    变换集合集成模块，用于将所述基于调整偏差的子空间近似变换集合集成到所述视频编码器和/或所述视频解码器的变换集合中。
  6. 根据权利要求5所述的变换集合生成装置，其中，所述最优预测残差块获取模块包括：
    预测残差块获取单元，用于获取所述原始视频帧的被编码单元在各个预测模式下的各个预测残差块，所述原始视频帧包括多个所述被编码单元；
    最优预测残差块确定单元，用于将所述被编码单元对应的各个预测残差块中函数值最小的预测残差块确定为所述被编码单元对应的最优预测残差块。
  7. 根据权利要求5或6所述的变换集合生成装置，其中，所述变换集合获取模块包括：
    残差块集合单元，用于将所述多个最优预测残差块划分并归类到多个最优预测残差块集合中；
    变换获取单元，用于根据所述多个最优预测残差块集合获取至少一个基于调整偏差的子空间近似变换；
    变换集合单元，用于将所述至少一个基于调整偏差的子空间近似变换划分 并归类到至少一个基于调整偏差的子空间近似变换集合中。
  8. 根据权利要求7所述的变换集合生成装置，其中，所述最优预测残差块集合中的最优预测残差块与集合聚类平均残差块的距离最小，所述集合聚类平均残差块是所述最优预测残差块集中所有最优预测残差块的平均。
  9. 一种机器可读存储介质，其存储有可执行指令，其中，所述指令当被执行时使得所述机器执行如权利要求1至4中任一所述的视频预测残差块的变换集合生成方法。
  10. 一种机器设备，其中，包括：
    至少一个处理器，以及
    与所述至少一个处理器耦合的存储器，所述存储器存储指令，当所述指令被所述至少一个处理器执行时，使得所述至少一个处理器执行如权利要求1至4中任一所述的视频预测残差块的变换集合生成方法。
PCT/CN2020/126033 2020-09-30 2020-11-03 变换集合生成方法和装置、机器可读存储介质和机器设备 Ceased WO2022067952A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011073947.7A CN114339217B (zh) 2020-09-30 2020-09-30 变换集合生成方法和装置、机器可读存储介质和机器设备
CN202011073947.7 2020-09-30

Publications (1)

Publication Number Publication Date
WO2022067952A1 true WO2022067952A1 (zh) 2022-04-07

Family

ID=80950995

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/126033 Ceased WO2022067952A1 (zh) 2020-09-30 2020-11-03 变换集合生成方法和装置、机器可读存储介质和机器设备

Country Status (2)

Country Link
CN (1) CN114339217B (zh)
WO (1) WO2022067952A1 (zh)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160219290A1 (en) * 2015-01-26 2016-07-28 Qualcomm Incorporated Enhanced multiple transforms for prediction residual
US20200021852A1 (en) * 2018-07-16 2020-01-16 Qualcomm Incorporated Transform variations of multiple separable transform selection

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160219290A1 (en) * 2015-01-26 2016-07-28 Qualcomm Incorporated Enhanced multiple transforms for prediction residual
US20200021852A1 (en) * 2018-07-16 2020-01-16 Qualcomm Incorporated Transform variations of multiple separable transform selection

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
C.-C. JAY KUO; MIN ZHANG; SIYANG LI; JIALI DUAN; YUERU CHEN: "Interpretable Convolutional Neural Networks via Feedforward Design", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 5 October 2018 (2018-10-05), 201 Olin Library Cornell University Ithaca, NY 14853 , XP080930258 *
LI NA; ZHANG YONGFEI; ZHANG YUN; KUO C.-C. JAY: "On Energy Compaction of 2D Saab Image Transforms", 2019 ASIA-PACIFIC SIGNAL AND INFORMATION PROCESSING ASSOCIATION ANNUAL SUMMIT AND CONFERENCE (APSIPA ASC), IEEE, 18 November 2019 (2019-11-18), pages 466 - 475, XP033733110, DOI: 10.1109/APSIPAASC47483.2019.9023199 *
ZHAO XIN; CHEN JIANLE; KARCZEWICZ MARTA; ZHANG LI; LI XIANG; CHIEN WEI-JUNG: "Enhanced Multiple Transform for Video Coding", 2016 DATA COMPRESSION CONFERENCE (DCC), IEEE, 30 March 2016 (2016-03-30), pages 73 - 82, XP033027689, DOI: 10.1109/DCC.2016.9 *

Also Published As

Publication number Publication date
CN114339217B (zh) 2024-09-13
CN114339217A (zh) 2022-04-12

Similar Documents

Publication Publication Date Title
RU2526764C2 (ru) Устройство, способ и программа для прогнозирующего кодирования изображений, устройство, способ и программа для прогнозирующего декодирования изображений и система и способ кодирования/декодирования
US10003792B2 (en) Video encoder for images
US9014499B2 (en) Distributed source coding using prediction modes obtained from side information
KR20110045908A (ko) 공간 분할을 이용한 움직임 벡터 부호화/복호화 방법 및 장치와 그를 이용한 영상 부호화/복호화 방법 및 장치
US8594189B1 (en) Apparatus and method for coding video using consistent regions and resolution scaling
Peng et al. Laplacian matrix learning for point cloud attribute compression with ternary search-based adaptive block partition
KR102261669B1 (ko) 인공신경망 기반 객체영역 검출방법, 장치 및 이에 대한 컴퓨터 프로그램
CN103020138A (zh) 一种视频检索的方法和装置
Li et al. Joint rate-distortion optimization for simultaneous texture and deep feature compression of facial images
Mei et al. Learn a compression for objection detection-vae with a bridge
CN105933711A (zh) 基于分割的邻域最优概率视频隐写分析方法及系统
CN111447438B (zh) 一种面向通用视频编码的快速帧内预测模式判决方法
KR102876564B1 (ko) 영상정보의 비트스트림정보에 기반하여 객체영역을 검출하는 방법 및 시스템
CN107231556B (zh) 一种图像云储存设备
CN112203087B (zh) 一种hevc帧内预测模式快速选择方法与系统
CN115834893A (zh) 视频编码方法及装置、并行计算单元及电子设备
WO2022067952A1 (zh) 变换集合生成方法和装置、机器可读存储介质和机器设备
Chen et al. CNN-optimized image compression with uncertainty based resource allocation
JP7364936B2 (ja) 符号化方法、符号化装置、及びプログラム
CN114025166A (zh) 视频压缩方法、电子设备及计算机可读存储介质
KR20100044333A (ko) 동영상 부호화 장치 및 이를 위한 영상 신호의 2차원 정렬 변환 장치 및 방법, 및 이를 위한 기록 매체
Zhang et al. Virtual Reference Frame-Based Inter Prediction for MPEG Enhanced G-PCC
CN115802048B (zh) 一种基于块间预测和图傅里叶变换的点云属性压缩方法
Ma et al. An adaptive lagrange multiplier determination method for dynamic texture in HEVC
US9307257B2 (en) Method for improving compression efficiency of distributed source coding using intra-band information

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20955972

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20955972

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20955972

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 11.12.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20955972

Country of ref document: EP

Kind code of ref document: A1