WO2023123478A1 - 预测方法、装置、设备、系统、及存储介质 - Google Patents

预测方法、装置、设备、系统、及存储介质 Download PDF

Info

Publication number
WO2023123478A1
WO2023123478A1 PCT/CN2021/143977 CN2021143977W WO2023123478A1 WO 2023123478 A1 WO2023123478 A1 WO 2023123478A1 CN 2021143977 W CN2021143977 W CN 2021143977W WO 2023123478 A1 WO2023123478 A1 WO 2023123478A1
Authority
WO
WIPO (PCT)
Prior art keywords
template
prediction
templates
mode
sub
Prior art date
Application number
PCT/CN2021/143977
Other languages
English (en)
French (fr)
Inventor
王凡
Original Assignee
Oppo广东移动通信有限公司
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 Oppo广东移动通信有限公司 filed Critical Oppo广东移动通信有限公司
Priority to PCT/CN2021/143977 priority Critical patent/WO2023123478A1/zh
Priority to CN202180105280.0A priority patent/CN118476224A/zh
Publication of WO2023123478A1 publication Critical patent/WO2023123478A1/zh

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
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction

Definitions

  • the present application relates to the technical field of video coding and decoding, and in particular to a prediction method, device, equipment, system, and storage medium.
  • Digital video technology can be incorporated into a variety of video devices, such as digital televisions, smartphones, computers, e-readers, or video players, among others.
  • video devices implement video compression technology to enable more effective transmission or storage of video data.
  • the prediction mode is determined first, for example, the first prediction mode and the second prediction mode of the current block are determined through template matching.
  • the division of the templates is not detailed enough at present, so that when the first prediction mode and the second prediction mode are determined according to the templates, the determined prediction mode is inaccurate, and thus the compression effect is poor.
  • Embodiments of the present application provide a prediction method, device, device, system, and storage medium, which improve the accuracy of template division and further improve compression performance.
  • the present application provides a prediction method applied to a decoder, including:
  • the embodiment of the present application provides a prediction method, including:
  • the present application provides a prediction device, configured to execute the method in the above first aspect or various implementation manners thereof.
  • the prediction device includes a functional unit configured to execute the method in the above first aspect or each implementation manner thereof.
  • the present application provides a prediction device, configured to execute the method in the above second aspect or various implementations thereof.
  • the prediction device includes a functional unit configured to execute the method in the above second aspect or each implementation manner thereof.
  • a video decoder including a processor and a memory.
  • the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory, so as to execute the method in the above first aspect or its various implementations.
  • a sixth aspect provides a video encoder, including a processor and a memory.
  • the memory is used to store a computer program
  • the processor is used to invoke and run the computer program stored in the memory, so as to execute the method in the above second aspect or its various implementations.
  • a video codec system including a video encoder and a video decoder.
  • the video decoder is configured to execute the method in the above first aspect or its various implementations
  • the video encoder is configured to execute the method in the above second aspect or its various implementations.
  • the chip includes: a processor, configured to call and run a computer program from the memory, so that the device installed with the chip executes any one of the above-mentioned first to second aspects or any of the implementations thereof. method.
  • a computer-readable storage medium for storing a computer program, and the computer program causes a computer to execute any one of the above-mentioned first to second aspects or the method in each implementation manner thereof.
  • a computer program product including computer program instructions, the computer program instructions cause a computer to execute any one of the above first to second aspects or the method in each implementation manner.
  • a computer program which, when running on a computer, causes the computer to execute any one of the above-mentioned first to second aspects or the method in each implementation manner thereof.
  • a code stream is provided, and the code stream is generated based on the method in the second aspect above.
  • the present application derives the mode based on the size and/or weight of the current block when determining K templates, so that the determined K templates are more in line with the actual situation, so that when using these K templates to determine the prediction mode, the prediction mode can be improved.
  • FIG. 1 is a schematic block diagram of a video encoding and decoding system involved in an embodiment of the present application
  • Fig. 2 is a schematic block diagram of a video encoder involved in an embodiment of the present application
  • Fig. 3 is a schematic block diagram of a video decoder involved in an embodiment of the present application.
  • Fig. 4 is a schematic diagram of weight distribution
  • Fig. 5 is a schematic diagram of weight distribution
  • FIG. 6A is a schematic diagram of inter-frame prediction
  • FIG. 6B is a schematic diagram of weighted inter prediction
  • FIG. 7A is a schematic diagram of intra prediction
  • FIG. 7B is a schematic diagram of intra prediction
  • 8A-8I are schematic diagrams of intra prediction
  • FIG. 9 is a schematic diagram of an intra prediction mode
  • FIG. 10 is a schematic diagram of an intra prediction mode
  • FIG. 11 is a schematic diagram of an intra prediction mode
  • FIG. 12 is a schematic diagram of weighted intra prediction
  • Figure 13 is a schematic diagram of template matching
  • FIG. 14 is a schematic flow chart of a prediction method provided by an embodiment of the present application.
  • Fig. 15 is a schematic diagram when using two prediction modes to predict the current block
  • Fig. 16 is a schematic diagram of template division
  • 17A-17G are schematic diagrams of template division
  • 18A-18D are schematic diagrams of another template division
  • Fig. 19 is a schematic diagram of template size
  • FIG. 20A is a schematic diagram of weight distribution
  • FIG. 20B is another schematic diagram of weight distribution
  • FIG. 21 is a schematic flow chart of a prediction method provided by an embodiment of the present application.
  • Fig. 22 is a schematic block diagram of a prediction device provided by an embodiment of the present application.
  • Fig. 23 is a schematic block diagram of a prediction device provided by an embodiment of the present application.
  • Fig. 24 is a schematic block diagram of an electronic device provided by an embodiment of the present application.
  • Fig. 25 is a schematic block diagram of a video encoding and decoding system provided by an embodiment of the present application.
  • the application can be applied to the field of image codec, video codec, hardware video codec, dedicated circuit video codec, real-time video codec, etc.
  • the solution of the present application can be combined with audio and video coding standards (audio video coding standard, referred to as AVS), for example, H.264/audio video coding (audio video coding, referred to as AVC) standard, H.265/high efficiency video coding ( High efficiency video coding (HEVC for short) standard and H.266/versatile video coding (VVC for short) standard.
  • the solutions of the present application may operate in conjunction with other proprietary or industry standards, including ITU-TH.261, ISO/IECMPEG-1Visual, ITU-TH.262 or ISO/IECMPEG-2Visual, ITU-TH.263 , ISO/IECMPEG-4Visual, ITU-TH.264 (also known as ISO/IECMPEG-4AVC), including scalable video codec (SVC) and multi-view video codec (MVC) extensions.
  • SVC scalable video codec
  • MVC multi-view video codec
  • FIG. 1 is a schematic block diagram of a video encoding and decoding system involved in an embodiment of the present application. It should be noted that FIG. 1 is only an example, and the video codec system in the embodiment of the present application includes but is not limited to what is shown in FIG. 1 .
  • the video codec system 100 includes an encoding device 110 and a decoding device 120 .
  • the encoding device is used to encode (can be understood as compression) the video data to generate a code stream, and transmit the code stream to the decoding device.
  • the decoding device decodes the code stream generated by the encoding device to obtain decoded video data.
  • the encoding device 110 in the embodiment of the present application can be understood as a device having a video encoding function
  • the decoding device 120 can be understood as a device having a video decoding function, that is, the embodiment of the present application includes a wider range of devices for the encoding device 110 and the decoding device 120, Examples include smartphones, desktop computers, mobile computing devices, notebook (eg, laptop) computers, tablet computers, set-top boxes, televisions, cameras, display devices, digital media players, video game consoles, vehicle-mounted computers, and the like.
  • the encoding device 110 may transmit the encoded video data (such as code stream) to the decoding device 120 via the channel 130 .
  • Channel 130 may include one or more media and/or devices capable of transmitting encoded video data from encoding device 110 to decoding device 120 .
  • channel 130 includes one or more communication media that enable encoding device 110 to transmit encoded video data directly to decoding device 120 in real-time.
  • encoding device 110 may modulate the encoded video data according to a communication standard and transmit the modulated video data to decoding device 120 .
  • the communication medium includes a wireless communication medium, such as a radio frequency spectrum.
  • the communication medium may also include a wired communication medium, such as one or more physical transmission lines.
  • the channel 130 includes a storage medium that can store video data encoded by the encoding device 110 .
  • the storage medium includes a variety of local access data storage media, such as optical discs, DVDs, flash memory, and the like.
  • the decoding device 120 may acquire encoded video data from the storage medium.
  • channel 130 may include a storage server that may store video data encoded by encoding device 110 .
  • the decoding device 120 may download the stored encoded video data from the storage server.
  • the storage server may store the encoded video data and may transmit the encoded video data to the decoding device 120, such as a web server (eg, for a website), a file transfer protocol (FTP) server, and the like.
  • FTP file transfer protocol
  • the encoding device 110 includes a video encoder 112 and an output interface 113 .
  • the output interface 113 may include a modulator/demodulator (modem) and/or a transmitter.
  • the encoding device 110 may include a video source 111 in addition to the video encoder 112 and the input interface 113 .
  • the video source 111 may include at least one of a video capture device (for example, a video camera), a video archive, a video input interface, a computer graphics system, wherein the video input interface is used to receive video data from a video content provider, and the computer graphics system Used to generate video data.
  • a video capture device for example, a video camera
  • a video archive for example, a video archive
  • a video input interface for example, a video archive
  • video input interface for example, a video input interface
  • computer graphics system used to generate video data.
  • the video encoder 112 encodes the video data from the video source 111 to generate a code stream.
  • Video data may include one or more pictures or a sequence of pictures.
  • the code stream contains the encoding information of an image or image sequence in the form of a bit stream.
  • the encoded information may include encoded image data and associated data.
  • the associated data may include a sequence parameter set (SPS for short), a picture parameter set (PPS for short) and other syntax structures.
  • SPS sequence parameter set
  • PPS picture parameter set
  • An SPS may contain parameters that apply to one or more sequences.
  • a PPS may contain parameters applied to one or more images.
  • the syntax structure refers to a set of zero or more syntax elements arranged in a specified order in the code stream.
  • the video encoder 112 directly transmits encoded video data to the decoding device 120 via the output interface 113 .
  • the encoded video data can also be stored on a storage medium or a storage server for subsequent reading by the decoding device 120 .
  • the decoding device 120 includes an input interface 121 and a video decoder 122 .
  • the decoding device 120 may include a display device 123 in addition to the input interface 121 and the video decoder 122 .
  • the input interface 121 includes a receiver and/or a modem.
  • the input interface 121 can receive encoded video data through the channel 130 .
  • the video decoder 122 is used to decode the encoded video data to obtain decoded video data, and transmit the decoded video data to the display device 123 .
  • the display device 123 displays the decoded video data.
  • the display device 123 may be integrated with the decoding device 120 or external to the decoding device 120 .
  • the display device 123 may include various display devices, such as a liquid crystal display (LCD), a plasma display, an organic light emitting diode (OLED) display, or other types of display devices.
  • LCD liquid crystal display
  • plasma display a plasma display
  • OLED organic light emitting diode
  • FIG. 1 is only an example, and the technical solutions of the embodiments of the present application are not limited to FIG. 1 .
  • the technology of the present application may also be applied to one-sided video encoding or one-sided video decoding.
  • Fig. 2 is a schematic block diagram of a video encoder involved in an embodiment of the present application. It should be understood that the video encoder 200 can be used to perform lossy compression on images, and can also be used to perform lossless compression on images.
  • the lossless compression may be visually lossless compression or mathematically lossless compression.
  • the video encoder 200 can be applied to image data in luminance-chrominance (YCbCr, YUV) format.
  • the YUV ratio can be 4:2:0, 4:2:2 or 4:4:4, Y means brightness (Luma), Cb (U) means blue chroma, Cr (V) means red chroma, U and V are expressed as chroma (Chroma) for describing color and saturation.
  • 4:2:0 means that every 4 pixels have 4 luminance components
  • 2 chroma components YYYYCbCr
  • 4:2:2 means that every 4 pixels have 4 luminance components
  • 4 Chroma component YYYYCbCrCbCr
  • 4:4:4 means full pixel display (YYYYCbCrCbCrCbCrCbCr).
  • the video encoder 200 reads video data, and divides a frame of image into several coding tree units (coding tree units, CTUs) for each frame of image in the video data.
  • CTB can be called “Tree block", “Largest Coding unit” (LCU for short) or “coding tree block” (CTB for short).
  • LCU Large Coding unit
  • CTB coding tree block
  • Each CTU may be associated with a pixel block of equal size within the image. Each pixel may correspond to one luminance (luma) sample and two chrominance (chrominance or chroma) samples.
  • each CTU may be associated with one block of luma samples and two blocks of chroma samples.
  • a CTU size is, for example, 128 ⁇ 128, 64 ⁇ 64, 32 ⁇ 32 and so on.
  • a CTU can be further divided into several coding units (Coding Unit, CU) for coding, and the CU can be a rectangular block or a square block.
  • the CU can be further divided into a prediction unit (PU for short) and a transform unit (TU for short), so that encoding, prediction, and transformation are separated, and processing is more flexible.
  • a CTU is divided into CUs in a quadtree manner, and a CU is divided into TUs and PUs in a quadtree manner.
  • the video encoder and video decoder can support various PU sizes. Assuming that the size of a specific CU is 2N ⁇ 2N, video encoders and video decoders may support 2N ⁇ 2N or N ⁇ N PU sizes for intra prediction, and support 2N ⁇ 2N, 2N ⁇ N, N ⁇ 2N, NxN or similarly sized symmetric PUs for inter prediction. The video encoder and video decoder may also support asymmetric PUs of 2NxnU, 2NxnD, nLx2N, and nRx2N for inter prediction.
  • the video encoder 200 may include: a prediction unit 210, a residual unit 220, a transform/quantization unit 230, an inverse transform/quantization unit 240, a reconstruction unit 250, and a loop filter unit 260. Decoded image cache 270 and entropy coding unit 280. It should be noted that the video encoder 200 may include more, less or different functional components.
  • the current block may be called a current coding unit (CU) or a current prediction unit (PU).
  • a predicted block may also be called a predicted image block or an image predicted block, and a reconstructed image block may also be called a reconstructed block or an image reconstructed image block.
  • the prediction unit 210 includes an inter prediction unit 211 and an intra estimation unit 212 . Because there is a strong correlation between adjacent pixels in a video frame, the intra-frame prediction method is used in video coding and decoding technology to eliminate the spatial redundancy between adjacent pixels. Due to the strong similarity between adjacent frames in video, the inter-frame prediction method is used in video coding and decoding technology to eliminate time redundancy between adjacent frames, thereby improving coding efficiency.
  • the inter-frame prediction unit 211 can be used for inter-frame prediction.
  • the inter-frame prediction can include motion estimation (motion estimation) and motion compensation (motion compensation). It can refer to image information of different frames.
  • the inter-frame prediction uses motion information to find reference frames from reference frames. Blocks, generate prediction blocks based on reference blocks to eliminate temporal redundancy; frames used for inter-frame prediction can be P frames and/or B frames, P frames refer to forward prediction frames, and B frames refer to bidirectional prediction frame.
  • Inter-frame prediction uses motion information to find a reference block from a reference frame, and generates a prediction block based on the reference block.
  • the motion information includes the reference frame list where the reference frame is located, the reference frame index, and the motion vector.
  • the motion vector can be an integer pixel or a sub-pixel.
  • the reference frame found according to the motion vector A block of whole pixels or sub-pixels is called a reference block.
  • Some technologies will directly use the reference block as a prediction block, and some technologies will further process the reference block to generate a prediction block. Reprocessing and generating a prediction block based on a reference block can also be understood as taking the reference block as a prediction block and then processing and generating a new prediction block based on the prediction block.
  • the intra-frame estimation unit 212 only refers to the information of the same frame of images to predict the pixel information in the current code image block for eliminating spatial redundancy.
  • a frame used for intra prediction may be an I frame.
  • the intra prediction modes used by HEVC include planar mode (Planar), DC and 33 angle modes, a total of 35 prediction modes.
  • the intra-frame modes used by VVC include Planar, DC and 65 angle modes, with a total of 67 prediction modes.
  • the intra-frame prediction will be more accurate, and it will be more in line with the demand for the development of high-definition and ultra-high-definition digital video.
  • the residual unit 220 may generate a residual block of the CU based on the pixel blocks of the CU and the prediction blocks of the PUs of the CU. For example, residual unit 220 may generate a residual block for a CU such that each sample in the residual block has a value equal to the difference between the samples in the pixel blocks of the CU, and the samples in the PUs of the CU. Corresponding samples in the predicted block.
  • Transform/quantization unit 230 may quantize the transform coefficients. Transform/quantization unit 230 may quantize transform coefficients associated with TUs of a CU based on quantization parameter (QP) values associated with the CU. Video encoder 200 may adjust the degree of quantization applied to transform coefficients associated with a CU by adjusting the QP value associated with the CU.
  • QP quantization parameter
  • Inverse transform/quantization unit 240 may apply inverse quantization and inverse transform to the quantized transform coefficients, respectively, to reconstruct a residual block from the quantized transform coefficients.
  • the reconstruction unit 250 may add samples of the reconstructed residual block to corresponding samples of one or more prediction blocks generated by the prediction unit 210 to generate a reconstructed image block associated with the TU. By reconstructing the sample blocks of each TU of the CU in this way, the video encoder 200 can reconstruct the pixel blocks of the CU.
  • the loop filtering unit 260 is used to process the inversely transformed and inversely quantized pixels, compensate for distortion information, and provide better references for subsequent encoded pixels. For example, deblocking filtering operations can be performed to reduce block effect.
  • the loop filtering unit 260 includes a deblocking filtering unit and a sample adaptive compensation/adaptive loop filtering (SAO/ALF) unit, wherein the deblocking filtering unit is used for deblocking, and the SAO/ALF unit Used to remove ringing effects.
  • SAO/ALF sample adaptive compensation/adaptive loop filtering
  • the decoded image buffer 270 may store reconstructed pixel blocks.
  • Inter prediction unit 211 may use reference pictures containing reconstructed pixel blocks to perform inter prediction on PUs of other pictures.
  • intra estimation unit 212 may use the reconstructed pixel blocks in decoded picture cache 270 to perform intra prediction on other PUs in the same picture as the CU.
  • Entropy encoding unit 280 may receive the quantized transform coefficients from transform/quantization unit 230 . Entropy encoding unit 280 may perform one or more entropy encoding operations on the quantized transform coefficients to generate entropy encoded data.
  • Fig. 3 is a schematic block diagram of a video decoder involved in an embodiment of the present application.
  • the video decoder 300 includes: an entropy decoding unit 310 , a prediction unit 320 , an inverse quantization/transformation unit 330 , a reconstruction unit 340 , a loop filter unit 350 and a decoded image buffer 360 . It should be noted that the video decoder 300 may include more, less or different functional components.
  • the video decoder 300 can receive code streams.
  • the entropy decoding unit 310 may parse the codestream to extract syntax elements from the codestream. As part of parsing the codestream, the entropy decoding unit 310 may parse the entropy-encoded syntax elements in the codestream.
  • the prediction unit 320 , the inverse quantization/transformation unit 330 , the reconstruction unit 340 and the loop filter unit 350 can decode video data according to the syntax elements extracted from the code stream, that is, generate decoded video data.
  • the prediction unit 320 includes an intra estimation unit 322 and an inter prediction unit 321 .
  • Intra estimation unit 322 may perform intra prediction to generate a predictive block for a PU. Intra-estimation unit 322 may use an intra-prediction mode to generate a predictive block for a PU based on pixel blocks of spatially neighboring PUs. Intra estimation unit 322 may also determine the intra prediction mode for the PU from one or more syntax elements parsed from the codestream.
  • the inter prediction unit 321 can construct the first reference picture list (list 0) and the second reference picture list (list 1) according to the syntax elements parsed from the codestream. Furthermore, if the PU is encoded using inter prediction, entropy decoding unit 310 may parse the motion information for the PU. Inter prediction unit 321 may determine one or more reference blocks for the PU according to the motion information of the PU. Inter prediction unit 321 may generate a prediction block for a PU based on one or more reference blocks of the PU.
  • Inverse quantization/transform unit 330 may inverse quantize (ie, dequantize) transform coefficients associated with a TU. Inverse quantization/transform unit 330 may use QP values associated with CUs of the TU to determine the degree of quantization.
  • inverse quantization/transform unit 330 may apply one or more inverse transforms to the inverse quantized transform coefficients in order to generate a residual block associated with the TU.
  • Reconstruction unit 340 uses the residual blocks associated with the TUs of the CU and the prediction blocks of the PUs of the CU to reconstruct the pixel blocks of the CU. For example, the reconstruction unit 340 may add the samples of the residual block to the corresponding samples of the prediction block to reconstruct the pixel block of the CU to obtain the reconstructed image block.
  • Loop filtering unit 350 may perform deblocking filtering operations to reduce blocking artifacts of pixel blocks associated with a CU.
  • Video decoder 300 may store the reconstructed picture of the CU in decoded picture cache 360 .
  • the video decoder 300 may use the reconstructed picture in the decoded picture buffer 360 as a reference picture for subsequent prediction, or transmit the reconstructed picture to a display device for presentation.
  • the basic process of video encoding and decoding is as follows: at the encoding end, a frame of image is divided into blocks, and for the current block, the prediction unit 210 uses intra-frame prediction or inter-frame prediction to generate a prediction block of the current block.
  • the residual unit 220 may calculate a residual block based on the predicted block and the original block of the current block, that is, a difference between the predicted block and the original block of the current block, and the residual block may also be referred to as residual information.
  • the residual block can be transformed and quantized by the transformation/quantization unit 230 to remove information that is not sensitive to human eyes, so as to eliminate visual redundancy.
  • the residual block before being transformed and quantized by the transform/quantization unit 230 may be called a time domain residual block, and the time domain residual block after being transformed and quantized by the transform/quantization unit 230 may be called a frequency residual block or a frequency-domain residual block.
  • the entropy coding unit 280 receives the quantized variation coefficients output by the variation quantization unit 230 , and may perform entropy coding on the quantized variation coefficients to output a code stream.
  • the entropy coding unit 280 can eliminate character redundancy according to the target context model and the probability information of the binary code stream.
  • the entropy decoding unit 310 can analyze the code stream to obtain the prediction information of the current block, the quantization coefficient matrix, etc., and the prediction unit 320 uses intra prediction or inter prediction for the current block based on the prediction information to generate a prediction block of the current block.
  • the inverse quantization/transformation unit 330 uses the quantization coefficient matrix obtained from the code stream to perform inverse quantization and inverse transformation on the quantization coefficient matrix to obtain a residual block.
  • the reconstruction unit 340 adds the predicted block and the residual block to obtain a reconstructed block.
  • the reconstructed blocks form a reconstructed image, and the loop filtering unit 350 performs loop filtering on the reconstructed image based on the image or based on the block to obtain a decoded image.
  • the encoding end also needs similar operations to the decoding end to obtain the decoded image.
  • the decoded image may also be referred to as a reconstructed image, and the reconstructed image may be a subsequent frame as a reference frame for inter-frame prediction.
  • the block division information determined by the encoder as well as mode information or parameter information such as prediction, transformation, quantization, entropy coding, and loop filtering, etc., are carried in the code stream when necessary.
  • the decoding end analyzes the code stream and analyzes the existing information to determine the same block division information as the encoding end, prediction, transformation, quantization, entropy coding, loop filtering and other mode information or parameter information, so as to ensure the decoding image obtained by the encoding end It is the same as the decoded image obtained by the decoder.
  • the above is the basic process of the video codec under the block-based hybrid coding framework. With the development of technology, some modules or steps of the framework or process may be optimized. This application is applicable to the block-based hybrid coding framework.
  • the basic process of the video codec but not limited to the framework and process.
  • the current block may be a current coding unit (CU), a current prediction unit (PU), or the like. Due to the need for parallel processing, images can be divided into slices, etc., and slices in the same image can be processed in parallel, that is to say, there is no data dependence between them. And "frame” is a commonly used term, which can generally be understood as a frame is an image. The frames mentioned in the application can also be replaced by images or slices, etc.
  • VVC Versatile Video Coding
  • ADP Angular Weighted Prediction
  • the traditional unidirectional prediction only finds a reference block with the same size as the current block
  • the traditional bidirectional prediction uses two reference blocks with the same size as the current block
  • the pixel value of each point in the predicted block is The average value of the corresponding positions of the two reference blocks, that is, all points of each reference block account for 50% of the proportion.
  • Bidirectional weighted prediction makes the proportions of two reference blocks different, for example, all points in the first reference block account for 75% of the proportion, and all points in the second reference block account for 25% of the proportion. But all points in the same reference block have the same scale.
  • DMVR Decoder sideMotion Vector Refinement
  • BIO bi-directional optical flow
  • GPM or AWP will also Use two reference blocks with the same size as the current block, but some pixel positions use 100% the pixel values corresponding to the first reference block, some pixel positions 100% use the pixel values corresponding to the second reference block, and In the boundary area, the pixel values of the corresponding positions of the two reference blocks are used according to a certain ratio.
  • GPM or AWP uses two reference blocks that are different in size from the current block, that is, each takes a required part as a reference block. That is, the part whose weight is not 0 is used as a reference block, and the part whose weight is 0 is eliminated.
  • FIG. 4 is a schematic diagram of weight distribution, as shown in FIG. 4 , which shows a schematic diagram of weight distribution of multiple division modes of a GPM on a 64 ⁇ 64 current block provided by an embodiment of the present application, wherein, There are 64 division modes in GPM.
  • Fig. 5 is a schematic diagram of weight distribution. As shown in Fig. 5, it shows a schematic diagram of weight distribution of various division modes of an AWP on a 64 ⁇ 64 current block provided by an embodiment of the present application, wherein there are 56 a division mode.
  • the black area indicates that the weight value of the corresponding position of the first reference block is 0%
  • the white area indicates that the weight value of the corresponding position of the first reference block is 100%
  • the gray area indicates that the weight value of the corresponding position of the first reference block is 100%.
  • the area indicates a weight value corresponding to the first reference block with a weight value greater than 0% and less than 100% according to the color depth, and the weight value corresponding to the second reference block is 100% minus the first The weight value of the corresponding position of a reference block.
  • GPM determines the angle and offset according to each mode, and then calculates the weight matrix of each mode.
  • AWP first makes a one-dimensional weight line, and then uses a method similar to intra-frame angle prediction to spread the one-dimensional weight line across the entire matrix.
  • GPM and AWP achieve the predicted non-rectangular division effect without division.
  • GPM and AWP use a mask of the weights of two reference blocks, ie the weight map mentioned above. This mask determines the weight of the two reference blocks when generating the prediction block, or it can be simply understood as part of the position of the prediction block comes from the first reference block and part of the position comes from the second reference block, and the transition area (blending area) weighted by the corresponding positions of the two reference blocks, so as to make the transition smoother.
  • GPM and AWP do not divide the current block into two CUs or PUs according to the dividing line, so the transformation, quantization, inverse transformation, and inverse quantization of the residual after prediction are also processed by taking the current block as a whole.
  • GPM simulates the division of geometry, more precisely the division of predictions, using weight matrices.
  • two predictors are required, and each predictor is determined by one unidirectional motion information. These two pieces of unidirectional motion information come from a motion information candidate list, for example, from a merge motion information candidate list (mergeCandList).
  • GPM uses two indexes in the code stream to determine 2 unidirectional motion information from mergeCandList.
  • Inter prediction uses motion information to represent "motion".
  • Basic motion information includes reference frame (reference frame) (or reference image (reference picture)) information and motion vector (MV, motion vector) information.
  • the commonly used bidirectional prediction uses two reference blocks to predict the current block. 2 reference blocks can use a forward reference block and a backward reference block. Optionally, both are forward or both are backward are allowed.
  • the so-called forward means that the time corresponding to the reference frame is before the current frame
  • the backward means that the time corresponding to the reference frame is after the current frame.
  • the forward direction refers to the position of the reference frame in the video before the current frame
  • the backward direction refers to the position of the reference frame in the video after the current frame.
  • the POC (picture order count) of the forward reference frame is smaller than the POC of the current frame
  • the POC of the backward reference frame is greater than the POC of the current frame.
  • two sets of reference frame information and motion vector information are required. Each of them can be understood as a one-way motion information, and the combination of these two groups forms a two-way motion information.
  • the unidirectional motion information and the bidirectional motion information can use the same data structure, but the two sets of reference frame information and the motion vector information of the bidirectional motion information are valid, and one of the reference frames of the unidirectional motion information information and motion vector information is invalid.
  • two reference frame lists are supported, denoted as RPL0 and RPL1, where RPL is an abbreviation for Reference Picture List.
  • RPL is an abbreviation for Reference Picture List.
  • the P slice can only use RPL0, and the B slice can use RPL0 and RPL1.
  • the codec finds a certain reference frame through the reference frame index.
  • motion information is represented by a reference frame index and a motion vector.
  • the reference frame index refIdxL0 corresponding to reference frame list 0 the motion vector mvL0 corresponding to reference frame list 0, the reference frame index refIdxL1 corresponding to reference frame list 1, and the motion corresponding to reference frame list 1 Vector mvL0.
  • the reference frame index corresponding to the reference frame list 0 and the reference frame index corresponding to the reference frame list 1 can be understood as the above-mentioned reference frame information.
  • two flag bits are used to respectively indicate whether to use the motion information corresponding to the reference frame list 0 and whether to use the motion information corresponding to the reference frame list 0, respectively marked as predFlagL0 and predFlagL1.
  • predFlagL0 and predFlagL1 indicate whether the above-mentioned one-way motion information is valid or not.
  • the data structure of motion information is not explicitly mentioned, it uses the reference frame index corresponding to each reference frame list, the motion vector and the "valid or not" flag to represent the motion information. In some standard texts, the motion information does not appear, but the motion vector is used. It can also be considered that the reference frame index and the flag of whether to use the corresponding motion information are attached to the motion vector. In this application, "motion information” is still used for convenience of description, but it should be understood that "motion vector” may also be used for description.
  • the motion information used by the current block can be saved.
  • Subsequent encoded and decoded blocks of the current frame can use motion information of previously encoded and decoded blocks, such as adjacent blocks, according to the adjacent positional relationship. This utilizes the correlation in the spatial domain, so this encoded and decoded motion information is called motion information in the spatial domain.
  • the motion information used by each block of the current frame can be preserved.
  • Subsequent codec frames can use the motion information of previous codec frames according to the reference relationship. This utilizes the correlation in the time domain, so the motion information of the encoded and decoded frames is called the motion information in the time domain.
  • the storage method of the motion information used by each block of the current frame usually uses a fixed-size matrix, such as a 4x4 matrix, as a minimum unit, and each minimum unit stores a group of motion information independently. In this way, every time a block is encoded and decoded, the smallest units corresponding to its position can store the motion information of this block. In this way, when using the motion information in the space domain or the motion information in the time domain, the motion information corresponding to the position can be found directly according to the position. If a 16x16 block uses traditional unidirectional prediction, then all 4x4 minimum units corresponding to this block store the motion information of this unidirectional prediction.
  • a fixed-size matrix such as a 4x4 matrix
  • a block uses GPM or AWP, then all the smallest units corresponding to this block will determine each smallest unit according to the GPM or AWP mode, the first motion information, the second motion information, and the position of each smallest unit Stored exercise information.
  • One method is that if all the 4x4 pixels corresponding to a minimum unit come from the first motion information, then this minimum unit stores the first motion information, and if all the 4x4 pixels corresponding to a minimum unit come from the second motion information , then this smallest unit stores the second motion information.
  • AWP will select one of the motion information for storage; the GPM approach is if the two motion information points to different references frame list, then combine them into two-way motion information storage, otherwise just store the second motion information.
  • the above mergeCandList is constructed according to spatial motion information, time domain motion information, history-based motion information, and some other motion information.
  • mergeCandList uses positions 1 to 5 in FIG. 6A to derive spatial motion information, and uses positions 6 or 7 in FIG. 6A to derive time domain motion information.
  • History-based motion information is to add the motion information of this block to a first-in-first-out list every time a block is encoded and decoded. The adding process may require some checks, such as whether it is duplicated with the existing motion information in the list. In this way, the motion information in this history-based list can be referred to when encoding and decoding the current block.
  • the syntax description about GPM is as shown in Table 1:
  • the current block may use CIIP or GPM. If the current block does not use CIIP, then it uses GPM, which is the content shown in the syntax "if(!ciip_flag[x0][y0])" in Table 1.
  • GPM needs to transmit three pieces of information in the code stream, namely merge_gpm_partition_idx, merge_gpm_idx0, and merge_gpm_idx1.
  • x0, y0 are used to determine the coordinates (x0, y0) of the luminance pixel in the upper left corner of the current block relative to the luminance pixel in the upper left corner of the image.
  • merge_gpm_partition_idx determines the partition shape of the GPM, as shown above, it is "analog partition", and merge_gpm_partition_idx is the weight derivation mode or the index of the weight derivation mode mentioned in the embodiment of this application.
  • merge_gpm_idx0 is the index value of the first motion information in the candidate list
  • merge_gpm_idx1 is the index value of the second motion information in the candidate list. If the candidate list length (MaxNumGpmMergeCand)>2, you need to pass merge_gpm_idx1, otherwise you can directly determine.
  • the decoding process of GPM includes the following steps:
  • the information input in the decoding process includes: the coordinates (xCb, yCb) of the brightness position of the upper left corner of the current block relative to the upper left corner of the image, the width cbWidth of the brightness component of the current block, the height cbHeight of the brightness component of the current block, and the brightness of 1/16 pixel accuracy
  • motion information may be represented by a combination of a motion vector, a reference frame index and a prediction list flag.
  • 2 reference frame lists are supported, each reference frame list may have multiple reference frames.
  • the unidirectional prediction uses only one reference block of one reference frame in one of the reference frame lists as a reference, and the bidirectional prediction uses one reference block of each reference frame in each of the two reference frame lists as a reference.
  • GPM uses 2 unidirectional forecasts.
  • a in the above mvA and mvB, mvCA and mvCB, refIdxA and refIdxB, predListFlagA and predListFlagB can be understood as the first prediction mode, and B can be understood as the second prediction mode.
  • predListFlagX indicates whether X uses the first reference frame list or the second reference frame list
  • refIdxX indicates the reference frame index in the reference frame list used by X
  • mvX indicates the brightness motion used by X Vector
  • mvCX represents the chroma motion vector used by X.
  • the information output by the decoding process includes: (cbWidth)X(cbHeight) luminance prediction sample matrix predSamplesL; (cbWidth/SubWidthC)X(cbHeight/SubHeightC) Cb chrominance component prediction sample matrix, if necessary; (cbWidth/SubWidthC) Prediction sample matrix for the Cr chroma component of X(cbHeight/SubHeightC), if required.
  • the luma component is used as an example below, and the processing of the chrominance component is similar to that of the luma component.
  • predSamplesLAL and predSamplesLBL are (cbWidth)X(cbHeight), which are prediction sample matrices made according to two prediction modes.
  • predSamplesL is derived as follows: predSamplesLAL and predSamplesLBL are determined according to luma motion vectors mvA and mvB, chrominance motion vectors mvCA and mvCB, reference frame indices refIdxA and refIdxB, and prediction list flags predListFlagA and predListFlagB, respectively. That is, the prediction is performed according to the motion information of the two prediction modes respectively, and the detailed process will not be repeated here.
  • GPM is a merge mode, and it can be considered that the two prediction modes of GPM are both merge modes.
  • nCbW is set to cbWidth
  • nCbH is set to cbHeight
  • the prediction sample matrices predSamplesLAL and predSamplesLBL made by the two prediction modes, and angleIdx and distanceIdx are used as input.
  • the weighted forecast derivation process of GPM includes the following steps:
  • the input of this process is: the width nCbW of the current block, the height nCbH of the current block; two (nCbW)X(nCbH) prediction sample matrices predSamplesLA and predSamplesLB; the "division" angle index variable angleIdx of GPM; the distance index variable of GPM distanceIdx; component index variable cIdx.
  • This example uses luminance as an example, so the cIdx above is 0, indicating a luminance component.
  • the output of this process is: (nCbW)X(nCbH) GPM prediction sample matrix pbSamples.
  • nW, nH, shift1, offset1, displacementX, displacementY, partFlip and shiftHor are derived as follows:
  • offsetY ((-nH)>>1)+(angleIdx ⁇ 16? (distanceIdx*nH)>>3:-((distanceIdx*nH)>>3)).
  • offsetX ((-nW)>>1)+(angleIdx ⁇ 16? (distanceIdx*nW)>>3:-((distanceIdx*nW)>>3),
  • variable wValue representing the weight of the prediction sample at the current position is derived as follows, wValue is the weight of the prediction value predSamplesLA[x][y] of the prediction matrix of the first prediction mode at point (x, y), and (8-wValue) That is, the weight of the predicted value predSamplesLB[x][y] of the prediction matrix of the first prediction mode at point (x, y).
  • the distance matrix disLut is determined according to Table 3:
  • weightIdx (((xL+offsetX) ⁇ 1)+1)*disLut[displacementX]+(((yL+offsetY) ⁇ 1)+1)*disLut[displacementY],
  • weightIdxL partFlip? 32+weightIdx:32–weightIdx,
  • pbSamples[x][y] Clip3(0,(1 ⁇ BitDepth)-1,(predSamplesLA[x][y]*wValue+predSamplesLB[x][y]*(8-wValue)+offset1)>> shift1).
  • a weight value is derived for each position of the current block, and then a predicted value pbSamples[x][y] of a GPM is calculated. Because this way the weight wValue does not have to be written in the form of a matrix, but it can be understood that if the wValue of each position is saved in a matrix, then it is a weight matrix. Calculate the weight of each point separately and weight it to get the predicted value of GPM, or calculate all the weights and then weight them uniformly to get the predicted sample matrix of GPM. The principle is the same.
  • the use of the weight matrix in many descriptions in this application is to make the expression easier to understand, and it is more intuitive to use the weight matrix to draw pictures. In fact, it can also be described according to the weight of each position.
  • the weight matrix export mode can also be called the weight export mode.
  • the decoding process of GPM can be expressed as: analyze the code stream, determine whether the current block uses GPM technology; if the current block uses GPM technology, determine the weight derivation mode (or "division" mode or weight matrix derivation mode), and the first motion information and the second motion information. Determine the first prediction block according to the first motion information, determine the second prediction block according to the second motion information, determine the weight matrix according to the weight matrix derivation mode, and determine the prediction of the current block according to the first prediction block and the second prediction block and the weight matrix piece.
  • GPM or AWP belongs to a prediction technology, and GPM or AWP needs to transmit a flag (flag) of whether GPM or AWP is used in the code stream, and the flag can indicate whether the current block is Use GPM or AWP.
  • the encoder needs to transmit the specific mode used in the code stream, that is, one of the 64 division modes of GPM, or one of the 56 division modes of AWP; and the index value of two unidirectional motion information. That is to say, for the current block, the decoder can obtain information about whether GPM or AWP is used by parsing the code stream.
  • the decoder can parse out the prediction mode parameters of GPM or AWP and two motion information Index value, for example, the current block can be divided into two partitions, then the first index value corresponding to the first partition and the second index value corresponding to the second partition can be parsed out.
  • the prediction mode parameters under GPM will be transmitted in the code stream, such as the specific division mode of GPM; usually, GPM includes 64 division modes.
  • GPM the specific division mode of GPM
  • AWP the prediction mode parameters under AWP will be transmitted in the code stream, such as the specific division mode of AWP; usually, AWP includes 56 division modes.
  • the current implementation method is to construct a unidirectional motion information candidate list on the encoder side by using the relevant information of the coded/decoded part before the current block, select unidirectional motion information from the unidirectional motion information candidate list, and combine the two unidirectional motion information Write the code stream to the index value (index) of the motion information in the unidirectional motion information candidate list.
  • the same method is adopted on the decoder side, that is, a unidirectional motion information candidate list is constructed using the relevant information of the decoded part before the current block, and the unidirectional motion information candidate list must be the same as the candidate list constructed on the encoder side. In this way, the index values of the two unidirectional motion information are parsed from the code stream, and then the two unidirectional motion information are found out from the unidirectional motion information candidate list, which is the two unidirectional motion information to be used by the current block.
  • the unidirectional motion information described in this application may include: motion vector information, that is, the value of (x, y), and corresponding reference frame information, that is, the reference frame list and the reference frame index in the reference frame list value.
  • motion vector information that is, the value of (x, y)
  • corresponding reference frame information that is, the reference frame list and the reference frame index in the reference frame list value.
  • One representation is to record the reference frame index values of two reference frame lists, where the reference frame index value corresponding to one reference frame list is valid, such as 0, 1, 2, etc.; the reference frame index value corresponding to the other reference frame list is Invalid, i.e. -1.
  • the reference frame list with valid reference frame index value is the reference frame list used by the motion information of the current block, and the corresponding reference frame can be found from the reference frame list according to the reference frame index value.
  • Each reference frame list has a corresponding motion vector, the motion vector corresponding to the valid reference frame list is valid, and the motion vector corresponding to the invalid reference frame list is invalid.
  • the decoder can find the required reference frame through the reference frame information in the unidirectional motion information, and can find the reference block in the reference frame according to the position of the current block and the value of the motion vector (x, y), and then determine the current block The inter-frame prediction value of .
  • the intra-frame prediction method uses coded and decoded reconstructed pixels surrounding the current block as reference pixels to predict the current block.
  • Figure 7A is a schematic diagram of intra-frame prediction. As shown in Figure 7A, the size of the current block is 4x4, and the pixels on the left row and upper column of the current block are reference pixels of the current block, and intra-frame prediction uses these reference pixels to predict the current block .
  • These reference pixels may all be available, that is, all have been encoded and decoded. Some parts may also be unavailable, for example, the current block is the leftmost of the whole frame, then the reference pixel on the left of the current block is unavailable.
  • the lower left part of the current block has not been encoded and decoded, so the reference pixel at the lower left is also unavailable.
  • the available reference pixel or some value or some method can be used for filling, or no filling is performed.
  • FIG. 7B is a schematic diagram of intra prediction.
  • the multiple reference line intra prediction method (Multiple reference line, MRL) can use more reference pixels to improve the encoding and decoding efficiency, for example, using 4 reference lines/ is listed as the reference pixel of the current block.
  • FIG. 8A-5I are schematic diagrams of intra-frame prediction.
  • intra-frame prediction for 4x4 blocks in H.264 mainly includes 9 modes. Among them, mode 0 as shown in FIG. 8A copies the pixels above the current block to the current block in the vertical direction as the prediction value, and mode 1 as shown in FIG. 8B copies the reference pixel on the left to the current block in the horizontal direction as the prediction value.
  • the mode 2 direct current DC shown in Figure 8C uses the average value of the 8 points A ⁇ D and I ⁇ L as the predicted value of all points, and the modes 3 ⁇ 8 shown in Figure 8D-5I respectively press a certain angle Copy the reference pixel to the corresponding position of the current block, because some positions of the current block cannot exactly correspond to the reference pixel, it may be necessary to use the weighted average of the reference pixel, or the sub-pixel of the interpolated reference pixel.
  • FIG. 9 is a schematic diagram of intra-frame prediction modes.
  • the intra-frame prediction modes used by HEVC include Planar, DC and 33 angle modes, a total of 35 prediction modes.
  • FIG. 10 is a schematic diagram of an intra-frame prediction mode.
  • the intra-frame modes used by VVC include Planar, DC and 65 angle modes, a total of 67 prediction modes.
  • Fig. 11 is a schematic diagram of intra prediction modes. As shown in Fig. 11, AVS3 uses DC, Planar, Bilinear and 63 angle modes, a total of 66 prediction modes.
  • the multiple intraprediction filter (MIPF) in AVS3 uses different filters to generate prediction values for different block sizes. For pixels at different positions in the same block, a filter is used to generate a prediction value for pixels that are closer to the reference pixel, and another filter is used to generate a prediction value for pixels far from the reference pixel.
  • MIPF multiple intraprediction filter
  • a technique for filtering predicted pixels, such as intraprediction filter (IPF) in AVS3, can use reference pixels to filter predicted values.
  • the most probable mode list (MostprobableModes List, MPM) intra-mode encoding technology can be used to improve the encoding and decoding efficiency.
  • MPM most probable mode list
  • intra-frame prediction mode of the surrounding encoded and decoded blocks and the intra-frame prediction mode derived from the intra-frame prediction mode of the surrounding encoded and decoded blocks, such as adjacent modes, and some intra-frame predictions that are commonly used or have a relatively high probability of use Modes, such as DC, Planar, Bilinear mode, etc., constitute a mode list.
  • Intra prediction modes that refer to surrounding coded blocks take advantage of spatial correlation. Because the texture will have a certain continuity in space. MPM can be used as the prediction of the intra prediction mode. That is to say, the probability that the current block uses MPM is higher than that of not using MPM. Therefore, during binarization, fewer codewords will be used for MPM, thereby saving overhead and improving encoding and decoding efficiency.
  • GPM combines two inter prediction blocks with a weight matrix. In fact it can be extended to combine two arbitrary prediction blocks. Such as two inter prediction blocks, two intra prediction blocks, one inter prediction block and one intra prediction block. Even in screen content coding, IBC (intra block copy) or palette prediction blocks can be used as one or two prediction blocks.
  • IBC intra block copy
  • palette prediction blocks can be used as one or two prediction blocks.
  • the prediction mode can be understood as information according to which the codec can generate a prediction block of the current block.
  • the prediction mode may be a certain intra-frame prediction mode, such as DC, Planar, various intra-frame angle prediction modes, and the like.
  • some or some auxiliary information can also be superimposed, such as the optimization method of the reference pixel in the frame, the optimization method (such as filtering) after the preliminary prediction block is generated, and the like.
  • the prediction mode can be skip (skip) mode, merge (merge) mode or MMVD (merge with motion vector difference, merge with motion vector difference) mode, or ordinary inter mode (MVP+ MVD), which can be one-way prediction or two-way prediction or multi-hypothesis prediction.
  • the inter-frame prediction mode uses unidirectional prediction, it must be able to determine a motion information, which is a unidirectional motion information, and the prediction block can be determined according to the motion information.
  • the inter-frame prediction mode uses bidirectional prediction, it must be able to determine one bidirectional motion information or two unidirectional motion information, and the prediction block can be determined according to the motion information.
  • the inter-frame prediction mode uses multi-hypothesis prediction, it must be able to determine multiple unidirectional motion information, and the prediction block can be determined according to the motion information.
  • skip, merge, and common inter mode can all support unidirectional prediction, bidirectional prediction or multi-hypothesis prediction.
  • a prediction mode is an inter-frame prediction mode, it can determine motion information, and a prediction block can be determined according to the motion information.
  • the template matching method can be used on the basis of skip mode and merge mode, MMVD mode, and ordinary inter mode. Such a prediction mode can still be called skip mode and merge mode, MMVD mode, ordinary inter mode or skip using template matching. mode, the merge mode using template matching, the MMVD mode using template matching, and the normal inter mode using template matching.
  • MMVD can be considered as a special merge mode, which indicates some specific MVDs through some flag bits, and these specific MVDs have only several possible preset values.
  • An example is the MMVD mode in VVC, which uses mmvd_direction_idx to indicate the direction of the MVD, and the possible values of mmvd_direction_idx are 0, 1, 2, 3.
  • mmvd_distance_idx 0 indicates that the horizontal component of MMVD is positive, and the vertical direction is 0; 1 indicates that of MMVD The horizontal component is negative, and the vertical direction is 0; 2 means that the horizontal component of MMVD is 0, and the vertical direction is positive; 3 means that the horizontal component of MMVD is 0, and the vertical direction is negative.
  • Use mmvd_distance_idx to represent the absolute value of the positive or negative value above.
  • the possible values of mmvd_distance_idx are 0 to 7.
  • the MVD of the ordinary inter mode can theoretically represent any possible MVD within a valid range.
  • the information that GPM needs to determine can be expressed as one weight derivation mode and two prediction modes.
  • the weight derivation mode is used to determine the weight matrix or weight, and the two prediction modes respectively determine a prediction block or a prediction value.
  • Weight export mode is also called partition mode in some places. But because it is an analog division, this application is more commonly called the weight derivation mode.
  • the two prediction modes may come from the same or different prediction methods, where the prediction methods include but not limited to intra prediction, inter prediction, IBC, and palette.
  • a specific concrete example is as follows: If the current block uses GPM. This example is used in inter-coded blocks, allowing the use of merge mode in intra prediction and inter prediction. As shown in Table 4, add a syntax element intra_mode_idx to indicate which prediction mode is an intra prediction mode.
  • intra_mode_idx 0
  • intra_mode_idx 1
  • intra_mode_idx 1
  • intra_mode_idx 2
  • intra_mode_idx 2
  • intra_mode_idx 3
  • both prediction modes are intra prediction modes, that is, mode0IsInter is 0, and mode0IsInter is 0.
  • the decoding process of GPM can be expressed as: analyze the code stream, determine whether the current block uses GPM technology; if the current block uses GPM technology, determine the weight derivation mode (or "division" mode or weight matrix derivation mode), and the first intra prediction mode and the second intra prediction mode.
  • the first prediction block is determined according to the first intra prediction mode
  • the second prediction block is determined according to the second intra prediction mode
  • the weight matrix is determined according to the weight matrix derivation mode
  • the weight matrix is determined according to the first prediction block and the second prediction block and the weight matrix The predicted block for the current block.
  • template matching was first used in inter-frame prediction, which uses the correlation between adjacent pixels to use some areas around the current block as templates.
  • inter-frame prediction uses the correlation between adjacent pixels to use some areas around the current block as templates.
  • the left and upper sides of the current block have been encoded and decoded according to the encoding order.
  • the existing hardware decoder is implemented, it may not be guaranteed that when the current block starts decoding, its left and upper sides have been decoded.
  • the inter-frame block such as the inter-frame coded block in HEVC.
  • the surrounding reconstructed pixels are not needed when predicting a block, so the prediction process of an inter block can be performed in parallel.
  • the intra-coded block must require the reconstructed pixels on the left side and the upper side as reference pixels.
  • the left side and the upper side are available, that is to say, it is achievable to adjust the hardware design accordingly.
  • the right side and the bottom side are not available under the encoding order of current standards such as VVC.
  • the left and upper rectangular areas of the current block are set as templates.
  • the height of the left template part is generally the same as the height of the current block, and the width of the upper template part is generally the same as the width of the current block. The same, of course, can also be different.
  • the so-called matching degree can be measured by some distortion costs, such as SAD (sum of absolute difference), SATD (sum of absolute transformed difference ), the transformation used by SATD is Hadamard transformation, MSE (mean-square error), etc.
  • SAD sum of absolute difference
  • SATD sum of absolute transformed difference
  • MSE mean-square error
  • the method of template matching may not be applicable to all blocks, so some methods can be used to determine whether the current block uses the above method of template matching, such as using a control switch in the current block to indicate whether to use template matching.
  • a name for this template matching method is DMVD (decoder side motion vector derivation).
  • Both the encoder and the decoder can use templates to search to derive motion information or find better motion information based on the original motion information. And it does not need to transmit specific motion vectors or motion vector differences, but both the encoder and the decoder perform the same rule search to ensure the consistency of encoding and decoding.
  • the method of template matching can improve the compression performance, but it also needs to "search" in the decoder, which brings a certain complexity of the decoder.
  • the method of template matching can also be used in intra frames, for example, using templates to determine the intra prediction mode.
  • the current block you can also use the area within a certain range on the top and left of the current block as a template, for example, the left rectangular area and the upper rectangular area as shown in the figure above.
  • the reconstructed pixels in the template are available when encoding and decoding the current block.
  • This process can roughly be described as determining a set of candidate intra-frame prediction modes for the current block, and the candidate intra-frame prediction modes constitute a subset of all available intra-frame prediction modes.
  • the candidate intra-frame prediction mode may be a complete set of all available intra-frame prediction modes.
  • the set of candidate intra prediction modes can be determined according to the MPM or some rules, such as equidistant screening. Calculate the cost of each candidate intra prediction mode on the template, such as SAD, SATD, MSE, etc. Use this mode to predict on the template to make a prediction block, and use the prediction block and the reconstruction block of the template to calculate the cost. A mode with a low cost may better match the template. Using the similarity between adjacent pixels, the intra prediction mode that performs well on the template may also be the intra prediction mode that performs well on the current block. Select one or several low-cost models. Of course, the above two steps can be repeated.
  • the set of candidate intra-frame prediction modes After selecting one or several low-cost modes, determine the set of candidate intra-frame prediction modes again, and then calculate the cost for the newly determined set of candidate intra-frame prediction modes. , select one or several low-cost models. This can also be understood as rough selection and fine selection.
  • the finally selected intra-frame prediction mode is determined as the intra-frame prediction mode of the current block, or several finally selected intra-frame prediction modes are used as candidates for the intra-frame prediction mode of the current block.
  • template matching such as sorting the MPM list, that is, the modes in the MPM list make prediction blocks on the template and determine the cost, from small to large Sort.
  • the higher the pattern in the MPM list the smaller the overhead in the code stream, which can also achieve the purpose of improving compression efficiency.
  • the method of template matching can be used to determine the two prediction modes of GPM. If the template matching method is used for GPM, one control switch can be used for the current block to control whether the two prediction modes of the current block use template matching, or two control switches can be used to control whether the two prediction modes use template matching respectively .
  • Another aspect is how to use template matching. For example, if GPM is used in merge mode, such as GPM in VVC, it uses merge_gpm_idxX to determine a motion information from mergeCandList, where the uppercase X is 0 or 1.
  • merge_gpm_idxX For the Xth motion information, one method is to use template matching method to optimize on the basis of the above motion information. That is, a motion information is determined from mergeCandList according to merge_gpm_idxX, and if template matching is used for the motion information, the template matching method is used to optimize on the basis of the above motion information.
  • Another method is not to use merge_gpm_idxX to determine a motion information from mergeCandList, but to search directly from a default motion information to determine a motion information.
  • the template matching method can be used to determine an intra-frame prediction mode, and there is no need to indicate the intra-frame prediction mode in the code stream index of. Or to determine a candidate set or MPM list by using a template matching method, it is necessary to indicate the index of the intra-frame prediction mode in the code stream.
  • the GPM can determine the area occupied by each prediction mode.
  • the so-called occupied area can be understood as the area whose weight corresponding to the prediction mode is the maximum value, or the area whose weight is greater than or equal to a certain threshold.
  • the reason why GPM can improve compression performance is because the two parts of the GPM "division" are different. Therefore, when using the template matching method to determine the prediction mode of the GPM, the template can also be divided.
  • the prior art can classify templates into 3 categories, namely left side, upper side and all (left side plus side).
  • the division of templates is related to the weight export mode. Exemplarily, as shown in Table 5, the template division in the prior art is related to the "division" angle or the "division" angle index angleIdx.
  • the template corresponding to the first prediction mode The template corresponding to the second prediction mode 0 TM_A TM_AL 1 / / 2 TM_A TM_AL 3 TM_A TM_AL 4 TM_A TM_L 5 TM_AL TM_L 6 / / 7 / / 8 TM_AL TM_L 9 / / 10 / / 11 TM_AL TM_L 12 TM_AL TM_AL 13 TM_A TM_AL 14 TM_A TM_AL 15 / / 16 TM_A TM_AL 17 / / 18 TM_A TM_AL 19 TM_A TM_AL 20 TM_A TM_L twenty one TM_AL TM_L twenty two / / twenty three / / twenty four TM_AL TM_L 25 / / 26 / / 27 TM_AL TM_L 28 TM_AL TM_AL 29 TM_A TM_AL
  • TM_A For example, record the left template as TM_A, the upper template as TM_L, and all (left plus side) templates as TM_AL.
  • Table 5 The relationship between the template and the "division" angle index is shown in Table 5. Some angle indexes such as 1, 6, and 7 are not used in the current GPM, so there is no corresponding template, which is represented by /.
  • Dividing the template according to the weight export mode does take into account the difference between the two parts of the GPM "division", but this division is actually not fine enough, because it only divides the template into two parts, the left side and the upper side, and according to the weight map. See the dividing line (the dividing line can be considered as a line composed of points with a median weight in the weight matrix. In the current GPM, the dividing line is a straight line. If there is no integer pixel point with a median weight in the actual weight matrix, then you can Sub-pixel points are used instead, of course, points of some other weight can also be used) may fall in various positions. In this way, when the template is selected according to the above Table 5, an inappropriate module may be selected, resulting in inaccurate prediction mode matching, which in turn leads to problems of low prediction accuracy and poor coding effect.
  • K templates are determined for the current block according to at least one of the current block size and weight derivation mode, and K prediction modes are determined using the K templates. That is to say, the present application is based on the size and/or weight derivation mode of the current block when determining the K templates, so that the determined K templates are more in line with the actual situation, and when using these K templates to determine the prediction mode, it can be improved.
  • the video decoding method provided by the embodiment of the present application will be introduced below with reference to FIG. 14 and by taking the decoding end as an example.
  • FIG. 14 is a schematic flowchart of a prediction method provided by an embodiment of the present application, and the embodiment of the present application is applied to the video decoder shown in FIG. 1 and FIG. 3 .
  • the method of the embodiment of the present application includes:
  • S101 Decode a code stream, and determine a weight derivation mode of a current block.
  • the weight derivation mode is used to determine the weight used by the current block.
  • the weight derivation mode may be a mode for deriving weights.
  • each weight export mode can export a weight matrix; for blocks of the same size, different weight export modes export different weight matrices.
  • AWP has 56 weight derivation modes
  • GPM has 64 weight derivation modes.
  • the ways for the decoder to determine the weight derivation mode of the current block include but are not limited to the following:
  • the decoding end selects the same weight derivation mode as the encoding end by default, for example, both the decoding end and the encoding end select the weight derivation mode with index 44.
  • the encoding end carries the index of the weight derivation mode used in the encoding process in the code stream. In this way, the decoding end can obtain the weight derivation mode of the current block by decoding the code stream.
  • Way 3 Determine the weight derivation mode in the same way as the encoding side.
  • the decoder tries all possible combinations of K prediction modes and weight derivation modes, K is a positive integer greater than 1, selects the weight derivation mode in the combination with the smallest cost, and determines it as the weight derivation mode of the current block.
  • the above K prediction modes include the first prediction mode and the second prediction mode, assuming that there are 66 available prediction modes, and the first prediction mode has 66 possibilities. Since the second prediction mode is different from the first The prediction modes are different, so there are 65 second prediction modes, assuming that there are 64 weight derivation modes (taking GPM as an example), then this application may use any two different prediction modes and any weight derivation mode, a total of 66 ⁇ 65 ⁇ 64 possibilities. If the PCM prediction mode is not used. Then there are 65 ⁇ 64 ⁇ 63 possibilities. It can be seen that in the present application, the selectable prediction modes and the number of usable weight derivation modes can also be limited, and the number of combinations will be correspondingly reduced.
  • the decoder may perform cost calculation on all possible combinations, and determine a combination with the smallest cost.
  • each combination includes a first prediction mode, a second prediction mode and a weight derivation mode.
  • all the above-mentioned possible combinations can be firstly selected, such as using SAD, SATD, etc. as the approximate cost for preliminary selection, and a set number of candidate first prediction modes,
  • SAD SAD
  • SATD SATD
  • more detailed cost calculation is performed to achieve fine selection, and a combination of the first prediction mode, the second prediction mode and the weight derivation mode with the smallest cost is determined. Therefore, some fast algorithms can be used to reduce the number of attempts during the primary selection. For example, when an angle prediction mode causes a high cost, several prediction modes adjacent to it will not be tried again.
  • the first prediction value will be determined according to the first prediction mode
  • the second prediction value will be determined according to the second prediction mode
  • the weight will be derived according to the weight derivation mode.
  • the predicted value, the second predicted value and the weight determine the predicted value for this application.
  • the SAD and SATD are determined by using the current block and the predicted value corresponding to the current block during primary selection of the SAD and SATD.
  • the above weight derived according to the weight derivation mode can be understood as deriving the weight corresponding to each pixel in the current block, and can also be understood as deriving the weight matrix corresponding to the current block.
  • the prediction value of the current block based on the weight when determining the prediction value of the current block based on the weight, it may be to determine the first prediction value and the second prediction value corresponding to each pixel in the current block, and according to the first prediction value and the second prediction value corresponding to each pixel value and weight to determine the prediction value corresponding to each pixel, and the prediction value corresponding to each pixel in the current block constitutes the prediction value of the current block.
  • the predicted value of the current block based on the weight it can also be performed according to the block. For example, the first predicted value and the second predicted value of the current block are determined, and the first predicted value of the current block is determined according to the weight matrix of the current block. The predicted value and the second predicted value are weighted to obtain a new predicted value of the current block.
  • the decoder before determining the weight derivation mode of the current block, the decoder first needs to determine whether the current block uses K different prediction modes for weighted prediction processing. If the decoder determines that the current block uses K different prediction modes for weighted prediction processing, it executes the above S101 to determine the weight derivation mode of the current block.
  • the decoding end may determine whether the current block uses K different prediction modes for weighted prediction processing by determining a prediction mode parameter of the current block.
  • the prediction mode parameter may indicate whether the current block can use GPM mode or AWP mode, that is, indicate whether the current block can use K different prediction modes for prediction processing.
  • the prediction mode parameter can be understood as a flag indicating whether the GPM mode or the AWP mode is used.
  • the encoder can use a variable as the prediction mode parameter, so that the setting of the prediction mode parameter can be realized by setting the value of the variable.
  • the encoder can set the value of the prediction mode parameter to indicate that the current block uses GPM mode or AWP mode.
  • the encoder can set the variable The value of is set to 1.
  • the encoder can set the value of the prediction mode parameter to indicate that the current block does not use GPM mode or AWP mode. Specifically, the encoder can Set the variable value to 0. Furthermore, in the embodiment of the present application, after the encoder finishes setting the prediction mode parameters, it can write the prediction mode parameters into the code stream and transmit them to the decoder, so that the decoder can Get the prediction mode parameters.
  • the decoder decodes the code stream to obtain the prediction mode parameters, and then determines whether the current block uses GPM mode or AWP mode according to the prediction mode parameters. If the current block uses GPM mode or AWP mode, it uses K different prediction modes. During prediction processing, the weight derivation mode for the current block is determined.
  • the GPM mode or the AWP mode is a prediction method, specifically, K different prediction modes are determined for the current block, and then determined according to the K different prediction modes. K predicted values, and then the weights can be determined again, and the K predicted values are combined according to the weights, and finally a new predicted value can be obtained.
  • the above K different prediction modes of the current block include the following examples:
  • Example 1 the above K different prediction modes are all intra-frame prediction modes.
  • Example 2 the above K different prediction modes are all inter-frame prediction modes.
  • Example 3 among the aforementioned K different prediction modes, at least one is an intra prediction mode, and at least one is an inter prediction mode.
  • Example 4 among the above K different prediction modes, at least one is an intra prediction mode, and at least one is a non-inter and non-intra prediction mode, such as an intra-block copy IBC prediction mode or a palette prediction mode, etc. .
  • Example 5 among the aforementioned K different prediction modes, at least one is an inter prediction mode, and at least one is a non-inter and non-intra prediction mode, for example, an IBC prediction mode or a palette prediction mode.
  • Example 6 none of the above K different prediction modes is an intra prediction mode or an inter prediction mode, for example, one is an IBC prediction mode, one is a palette prediction mode, and so on.
  • Fig. 15 is a schematic diagram of using two prediction modes to predict the current block. As shown in Fig. 15, when predicting the current block, the first prediction mode can be used to determine the first prediction value, while the second prediction mode can be used to determine The second predictive value can then use weights to combine the first predictive value and the second predictive value to finally obtain a new predictive value.
  • the size of the current block can be limited.
  • the decoder may first determine the size parameter of the current block, and then determine whether the current block uses the GPM mode or the AWP mode according to the size parameter.
  • the size parameter of the current block may include the height and width of the current block, therefore, the decoder may use the height and width of the current block to restrict the use of GPM mode or AWP mode.
  • the width is greater than the first threshold and the height is greater than the second threshold, it is determined that the current block uses the GPM mode or the AWP mode. It can be seen that a possible restriction is to use GPM mode or AWP mode only when the width of the block is greater than (or greater than or equal to) the first threshold and the height of the block is greater than (or greater than or equal to) the second threshold.
  • the values of the first threshold and the second threshold may be 8, 16, 32, etc., and the first threshold may be equal to the second threshold.
  • the width is smaller than the third threshold and the height is larger than the fourth threshold, it is determined that the current block uses the GPM mode or the AWP mode. It can be seen that a possible restriction is to use the SAWP mode only when the width of the block is smaller than (or smaller than or equal to) the third threshold and the height of the block is larger than (or larger than or equal to) the fourth threshold.
  • the values of the third threshold and the fourth threshold may be 8, 16, 32, etc., and the third threshold may be equal to the fourth threshold.
  • the limitation of the size of the block that can use the GPM mode or the AWP mode can also be realized through the limitation of the pixel parameter.
  • the decoder may first determine the pixel parameters of the current block, and then further judge whether the current block can use the GPM mode or the AWP mode according to the pixel parameters and the fifth threshold. It can be seen that a possible restriction is to use the GPM mode or the AWP mode only when the pixel number of the block is greater than (or greater than or equal to) the fifth threshold. Wherein, the value of the fifth threshold may be 8, 16, 32 and so on.
  • the current block can use the GPM mode or the AWP mode only if the size parameter of the current block meets the size requirement.
  • this application there may be a frame-level flag to determine whether the current frame to be decoded uses this application.
  • an intra frame such as an I frame
  • an inter frame such as a B frame, P frame
  • intra frames do not use this application
  • inter frames use this application.
  • Inter frames can also use intra prediction, thus inter frames are also likely to use this application.
  • a flag below the frame level and above the CU level such as tile, slice, patch, LCU, etc.
  • K is a positive integer greater than 1.
  • Template matching uses the correlation between adjacent pixels to use some areas around the current block as templates. When encoding and decoding the current block, its left side and upper side have been decoded according to the encoding order. During inter-frame prediction, the best matching position of the template is found in the reference frame to determine the motion information or motion vector of the current block. During intra-frame prediction, a template is used to determine the intra-frame prediction mode of the current block.
  • the present application does not limit the specific shape of the template of the current block.
  • the template of the current block includes at least one of an upper decoded region and a left decoded region of the current block.
  • the upper decoded area has the same width as the current block, and the left decoded area has the same height as the current block.
  • the templates corresponding to the first prediction mode and the second prediction mode are the upper decoded area of the current block, or the decoded area on the left side of the current block, or the decoded and left side of the current block.
  • the above decoded region the template corresponding to the first prediction mode is used to determine the first prediction mode
  • the template corresponding to the second prediction mode is used to determine the second prediction mode.
  • the white area of the weight matrix of the current block is the weight corresponding to the predicted value of the first prediction mode
  • the black area is the weight corresponding to the predicted value of the second prediction mode. the weight of.
  • the template corresponding to the first prediction mode is the upper decoded area of the current block
  • the template corresponding to the second prediction mode is the left decoded area of the current block
  • the templates close to the second prediction mode include In addition to the left area, part of the upper decoded area is also included. Therefore, the division of templates in the prior art is not fine enough, which leads to inaccurate determination of the prediction mode and large prediction errors when determining the prediction mode based on the imprecise template.
  • the embodiment of the present application implements fine division of templates by using at least one of the size of the current block and the weight derivation mode.
  • the process of determining K templates in S102 above based on at least one of the derived modes based on the size and weight of the current block will be described in detail below in conjunction with the methods proposed in Case 1 and Case 2 below.
  • the embodiment of the present application can implement a finer division of templates through the weight derivation mode.
  • the above S102 includes the following steps:
  • the template related to the second prediction mode includes not only the left area, but also the left part in the upper area, and the template related to the first prediction mode includes the right part in the upper area.
  • the matching accuracy of the second prediction mode can be improved when the left area and the left part of the upper area in the template of the current block are used as the template of the second prediction mode for the second prediction mode matching.
  • the matching accuracy of the first prediction mode can be improved. It can be seen that based on the weight derivation mode, accurate division of the template can be realized, thereby improving the accurate determination of the prediction mode and improving the decoding effect.
  • the ways of dividing the template of the current block into K templates include but are not limited to the following:
  • Way 1 divide the template of the current block into K templates according to the boundary line of the weight matrix corresponding to the weight derivation mode.
  • the present application extends the boundary line of the weight matrix corresponding to the weight derivation mode of the current block to the template of the current block to divide the template of the current block.
  • the boundary line can be extended to the right
  • the template on the side of the dividing line is recorded as the first template
  • the template on the left side of the dividing line is recorded as the second template.
  • the first template corresponds to the first prediction mode
  • the second template corresponds to the second prediction mode.
  • the first template can be used to derive the first prediction mode
  • the second template can be used to derive the second prediction mode, thereby realizing the prediction mode. Accurately determine and improve the decoding effect.
  • the first template and the second template divided according to the above method may not be rectangular.
  • the first template and the second template have hypotenuses, and the cost calculation for irregular templates is more complicated.
  • both the first template and the second template can be divided into rectangles.
  • the dividing line is extended to the template of the current block to obtain is used to divide the template of the current block into a first template and a second template, wherein the dividing line between the first template and the second template intersects with the extension line or does not intersect with the extension line.
  • the boundary line between the first template and the second template passes through an end point of the extension line and is perpendicular to the length side of the current block.
  • the boundary line between the first template and the second template passes through the midpoint of the extension line and is perpendicular to the length side of the current block.
  • the template of the current block is divided into K templates according to the dividing line of the weight matrix, which is simple and can realize accurate division of the templates.
  • the template of the current block may also be divided into K templates according to the second method as follows.
  • S102-A includes the following steps of S102-A1 and S102-A2:
  • the template of the current block is first divided into multiple sub-templates, for example, divided into M sub-templates, and then it is determined which template each sub-module corresponds to, and then the division of K templates is realized.
  • the embodiment of the present application does not limit the manner of dividing the foregoing sub-templates.
  • S102-A1 includes: dividing the template of the current block into M sub-templates according to the weight derivation mode.
  • Example 1 determine the weight matrix according to the weight derivation mode, extend the weight matrix to the template of the current block, for example, extend to the left and upward, and cover the weight matrix on the template of the current block. For example, as shown in FIG. 17D , you can choose to add the small rectangular area on the upper left side of the current block to the template of the current block, and combine the template of the current block and the current block to form a rectangle. Of course, it is also possible to use only the left part and the upper part as templates for the current block. As shown in FIG. 17D , the template of the current block includes the left area and the upper area of the current block, and the lower right rectangular area is the current block.
  • the template of the current block can be divided into M sub-templates according to the coverage of the template of the current block by the weight matrix.
  • the points whose weights are in the interval from a0 to a1 are divided into the first sub-template, and the points whose weights are in the interval from a1 to a2 are divided into the second sub-template, and so on.
  • the points in the interval from 1 to aM are divided into the Mth sub-template.
  • the black template in FIG. 17D is divided into the first sub-template, the upper gray template is divided into the second sub-template, and the upper white template is divided into the second sub-template.
  • the black template on the left side in FIG. Divided into a fourth sub-template.
  • the present application does not limit the specific shapes of the above M sub-templates.
  • the above example 1 divides the M sub-templates into rectangles.
  • Example 2 according to the weight derivation mode, determine the boundary line of the weight, and extend the boundary line to the template of the current block, so as to divide the template of the current block into M sub-templates.
  • the weight demarcation line is determined according to the weight derivation mode. From the description of the above embodiment, it can be seen that the demarcation line is a straight line formed by points whose weights change in the weight matrix composed of the weights of the points in the current block derived by the weight derivation mode. (or curve), such as the oblique line in Figure 17E.
  • the dividing line is extended to the template of the current block, and the upper template of the current block is divided into two parts. In this way, M sub-templates can be determined according to the templates divided by the weight dividing line. For example, as shown in FIG.
  • the template on the right side of the dividing line is divided into the first sub-template
  • the template on the left side of the dividing line is divided into the second sub-template
  • the template on the left side of the current block is divided into Divided into a third sub-template, at this time, the template of the current block is divided into three sub-templates.
  • the left template of the current block can also be divided into multiple sub-templates, for example, divided into two sub-modules. In this case, the template of the current block is divided into four sub-templates.
  • the template divided by the dividing line may be further divided according to other rules to obtain M sub-templates.
  • the first template and the second template divided according to the above method may not be rectangular.
  • the calculation cost of template matching is more complicated.
  • the boundary line is extended to the template of the current block to obtain the extension line of the boundary line in the template of the current block; using the extension line, the template of the current block is divided into M rectangular subtemplates.
  • the first sub-template and the second sub-template are divided into rectangles using extension lines.
  • the boundary line between the first sub-template and the second sub-template shown in FIG. 17G passes through the left endpoint of the extension line and is perpendicular to the length side of the current block.
  • the boundary line between the first sub-template and the second sub-template passes through The right endpoint of the extension line is perpendicular to the length side of the current block, or the boundary line between the first sub-template and the second sub-template passes through the midpoint of the extension line and is perpendicular to the length side of the current block.
  • the boundary line between the first sub-template and the second sub-template does not intersect the extension line, and is perpendicular to the length side of the current block.
  • method 2 in addition to dividing the template of the current block into M sub-templates according to the above-mentioned weight derivation mode, the following implementation method 2 can also be used to divide the template of the current block into M sub-templates, as shown below.
  • the template of the current block is divided into M sub-templates, that is, the above S102-A1 includes the following steps:
  • both P and Q are integers less than or equal to M, and the sum of P and Q is equal to M.
  • the template of the current block includes several rows of pixels that have been decoded above the current block and several columns of pixels that have been decoded on the left side of the current block. Several pixel rows are marked as the upper template of the current block, and several columns of pixels decoded on the left side of the current block are marked as the left template of the current block.
  • the template of the current block also includes the decoded area of the upper left corner of the current block, and/or includes the decoded area of the lower left of the current block, etc.
  • the embodiment of the present application does not limit the specific template of the current block.
  • the division of the upper template and the left template among the templates of the current block is mainly described as an example.
  • implementation mode 2 there is no limit to the way of dividing the upper template of the current block into P sub-templates and/or dividing the left template of the current block into Q sub-templates, for example, it can be divided equally or according to a preset ratio. Divide, or divide according to the preset number of pixels, or divide according to the preset number of pixel rows or pixel columns, etc.
  • the manners of dividing the left template of the current block into P sub-templates in the above S102-A11 include but are not limited to the following:
  • Mode 1 divide the upper template into P sub-templates along the vertical direction.
  • the upper template of the current block is evenly divided into P equal parts.
  • the upper template of the current block can be evenly divided into 3 equal parts, 4 equal parts, 5 equal parts, etc. That is to say, the embodiment of the present application does not limit the specific value of P, which can be determined according to actual needs. Sure.
  • the size of the P-1 sub-templates can be divided into the same size, and the size of the remaining sub-template is the same as the above-mentioned P-
  • the size of one sub-template is inconsistent, for example, the size of the remaining sub-template is smaller than the size of the above P-1 sub-templates, or the size of the remaining one sub-template is larger than the size of the above-mentioned P-1 sub-templates.
  • the upper template of the current block is divided into P sub-templates according to a preset ratio of sub-templates.
  • the ratio of a1:a2 the upper template of the current block is divided into two sub-templates.
  • the size ratio is 1:1.5.
  • the upper template of the current block is divided into three sub-templates.
  • a1:a2:a3 1:1.5:2, so according to the ratio of 1:1.5:2, the upper template of the current block is divided into 3 sub-templates, which are respectively recorded as sub-template 1, sub-template 2 and sub-template 3.
  • the size ratio of sub-template 1, sub-template 2 and sub-template 3 is 1:1.5:2.
  • Mode 2 divide the upper template into P sub-templates according to the preset number of pixels.
  • the preset number of pixels is used as a minimum division unit, and the upper template of the current block is divided into P sub-templates.
  • the present application does not limit the specific arrangement of the preset pixels, for example, the preset number of pixels is arranged into a rectangle, and this rectangular block is used as the minimum division unit of the upper template for division.
  • n columns of pixels are used as a minimum division unit, and the upper template is divided into P sub-templates, where n is a positive integer.
  • the template divided by each minimum division unit may be used as a sub-template.
  • each of the above four units is used as a sub-template to obtain two sub-templates. That is to say, in the embodiment of the present application, when n columns of pixels are used as the minimum division unit to divide the upper template of the current block, multiple templates divided by the minimum division unit can be used as a sub-template, for example, the minimum division unit is divided into Two or more adjacent areas of the template are used as a sub-template.
  • the present application does not limit the specific value of the above n, for example, it is a preset value.
  • the length of the upper template of the current block is the same as the length of the current block, so that the aforementioned n can be determined according to the length of the current block, for example, the length of the current block is a positive integer multiple of n.
  • the length of the current block is 16, the n may be 2, 4, 8 and other values.
  • the division method of the left template may be the same as or different from the division method of the upper template of the current block.
  • the manners of dividing the left template of the current block into Q sub-templates in the above S102-A11 include but are not limited to the following:
  • Mode 1 divide the left template into Q sub-templates along the horizontal direction.
  • the left template of the current block is equally divided into Q equal parts along the horizontal direction.
  • the template on the left side of the current block can be evenly divided into 3 equal parts, 4 equal parts, 5 equal parts, etc. That is to say, the embodiment of the present application does not limit the specific value of Q, and it will be determined according to the actual situation. Need to be sure.
  • the size of Q-1 sub-templates can be divided into the same size, and the size of the remaining sub-template is the same as the above-mentioned Q -
  • the sizes of the 1 sub-templates are inconsistent, for example, the size of the remaining sub-template is smaller than the size of the above Q-1 sub-templates, or the size of the remaining one sub-template is larger than the size of the above-mentioned Q-1 sub-templates.
  • the left template of the current block is divided into Q sub-templates according to a preset ratio of sub-templates.
  • the left template of the current block is divided into 2 sub-templates.
  • the size ratio is 1:1.5.
  • the left template of the current block is divided into three sub-templates.
  • b1:b2:b3 1:1.5:2, in this way, according to the ratio of 1:1.5:2, the left template of the current block is divided into 3 sub-templates, which are respectively recorded as sub-template 3, sub-template 4 and sub-template
  • the size ratio of template 5, sub-template 3, sub-template 4 and sub-template 5 is 1:1.5:2.
  • Mode 2 divide the left template into Q sub-templates according to the preset number of pixels.
  • the preset number of pixels is used as a minimum division unit, and the left template of the current block is divided into Q sub-templates.
  • the present application does not limit the specific arrangement of the preset pixels.
  • the preset number of pixels is arranged in a rectangle, and this rectangular block is used as the smallest division unit of the left template for division.
  • m rows of pixels are used as a minimum division unit, and the left template is divided into Q sub-templates, where m is a positive integer.
  • the left template of the current block can be
  • Each 4 rows of pixel rows in is divided into a unit, and then 4 units are obtained, and Q sub-templates are obtained according to these 4 units. For example, if each of the 4 units is used as a sub-template, 4 sub-templates are obtained.
  • the template divided by each minimum division unit may be used as a sub-template.
  • each of the above four units is used as a sub-template to obtain two sub-templates.
  • templates divided by multiple minimum division units can be used as a sub-template, for example, the minimum division unit The divided adjacent two or more regions are used as a sub-template.
  • the present application does not limit the specific value of the above m, for example, it is a preset value.
  • the width of the left template of the current block is the same as the width of the current block, so that the aforementioned m can be determined according to the width of the current block, for example, the width of the current block is a positive integer multiple of m.
  • the width of the current block is 16, then the m may be 2, 4, 8 and other values.
  • the above step S102-A2 is performed, that is, the M sub-templates are mapped to the K templates according to the weight derivation mode.
  • the template of the current block is divided into multiple sub-templates, for example, the template of the current block is divided into M sub-templates, and then, it is determined which sub-template each sub-template of the M sub-templates belongs to Templates, and then map M sub-templates to K templates to achieve fine and accurate division of templates.
  • the implementation methods of corresponding M sub-templates to K templates include but are not limited to the following:
  • Mode 1 according to the dividing line of the weight matrix, the M sub-templates are mapped to the K templates.
  • the sub-templates close to the first prediction mode are mapped to the first template
  • the sub-templates close to the second prediction mode are mapped to the second template.
  • the upper template of the current block is divided into four sub-templates, namely sub-template 1, sub-template 2, sub-template 3 and sub-template 4, and the left template of the current block is divided into two sub-templates, respectively Subtemplate 5 and Subtemplate 6.
  • sub-template 1 and sub-template 2 are close to the first prediction mode, therefore, sub-template 1 and sub-template 2 are corresponding to the first template, and sub-template 3, sub-template 4, sub-template 5 and sub-template 6 are close to the first prediction mode.
  • sub-template 3, sub-template 4, sub-template 5, and sub-template 6 are mapped to the second template.
  • the first template includes sub-template 1 and sub-template 2
  • sub-template 1 and sub-template 2 are used as templates to determine the first prediction mode of the current block.
  • the second template includes sub-template 3, sub-template 4, sub-template 5, and sub-template 6, and when the template is matched, sub-template 3, sub-template 4, sub-template 5, and sub-template 6 are used as templates to determine the first block of the current block. Two prediction modes, and then realize the accurate determination of the first prediction mode and the second prediction mode.
  • the sub-template can be mapped to the first template and the second template.
  • the first template and the second template have overlapping part.
  • the upper template of the current block is divided into four sub-templates, namely sub-template 1, sub-template 2, sub-template 3, and sub-template 4, and the left template of the current block is divided into two sub-templates, respectively Sub-template 5 and sub-template 6, and the dividing line of weight divides sub-template 3 into two parts, that is to say, the dividing line of weight intersects with sub-template 3.
  • the sub-template 3 can be as shown in FIG. 18C
  • the sub-template 3 is mapped to the second template, and the generated first template does not overlap with the second template.
  • the sub-template 3 can be respectively corresponding to the first template and the second template, that is, the first template includes sub-template 1, sub-template 2 and sub-template 3, and the second template includes Sub-template 3, sub-template 4, sub-template 5 and sub-template 6, now the first template and the second template have overlapping parts, so that when subsequent templates are matched, sub-template 3 can be used to determine the first prediction mode, and also It can be used to determine the second prediction mode, thereby enriching the template division methods.
  • the sub-template is mapped to the first template or the second template by default. For example, as shown in FIG. 18C , the dividing line of weight divides the sub-template 3 into two parts, and the sub-template is mapped to the second template by default.
  • the sub-template is corresponding to a smaller template, for example, as shown in Figure 18E, except for sub-template 3, the sub-template One template includes sub-template 1 and sub-template 2, and the second template includes sub-template 4, sub-template 5, and sub-template 6.
  • the area of the second template is much larger than the area of the first template.
  • the accuracy of template matching is improved as an example.
  • the sub-template 3 is corresponding to the first template with a smaller area, so as to increase the area of the first template, thereby improving the accuracy of determining the first prediction mode based on the first template.
  • the dividing line of the weight divides a sub-template into two parts. If the area of the sub-template in the first prediction mode is larger than the area in the second prediction mode, the sub-template is corresponding to the second prediction mode. In a template, for example as shown in FIG. 18C , the area of the sub-template 3 in the first prediction mode is larger than the area in the second prediction mode, so the sub-template 3 is corresponding to the first template. Optionally, if the area of the sub-template in the second prediction mode is larger than the area in the first prediction mode, then the sub-template is mapped to the second template.
  • the above S102-A2 includes the following steps:
  • the weight of the pixel points in the sub-template it is determined which template the sub-template is divided into. For example, the weight of the pixel points in the sub-template is the same or basically the same as the weight corresponding to the first prediction mode, then The sub-template corresponds to the first template, and if the weight of the pixel in the sub-template is the same or substantially the same as the weight corresponding to the second prediction mode, then the sub-template is corresponding to the second template.
  • the embodiment of the present application takes the j-th sub-template among the M sub-templates as an example to illustrate the process of determining which template other sub-templates correspond to Just refer to the jth sub-template.
  • determining a pixel point in the jth sub-template, such as the weight of the first point, and according to the weight of the first point determine which pixel the j-th sub-template corresponds to in the template.
  • the above-mentioned first point is any point in the jth submodule.
  • the above-mentioned first point is a point on the boundary line between the jth sub-template and the current block.
  • the method of determining the weight of the first point in the j-th template with respect to each of the K prediction modes is the same, and the embodiment of the present application takes the determination of the weight of the first point with respect to the i-th prediction mode as an example for illustration.
  • the way of determining the weight of the first point in the j-th sub-template with respect to the i-th prediction mode in S102-A21 above includes but is not limited to the following examples:
  • the weight matrix of the current block is extended to the j-th sub-template, so that the weight matrix of the current block at least covers the first point in the j-th sub-template, and then the weight of the first point is obtained.
  • the weight of the first point in the j-th sub-template with respect to the i-th prediction mode is determined, that is, the above-mentioned S102-A21 includes the following steps:
  • the weight of the first point in the j-th sub-template with respect to the i-th prediction mode is derived through the weight derivation mode.
  • the angle index and the distance index are determined according to the weight derivation mode, where the angle index can be understood as The angle index of the dividing line of each weight exported by the weight export mode.
  • the angle index and distance index corresponding to the weight derivation mode can be determined according to the above Table 2. For example, if the weight derivation mode is 27, the corresponding angle index is 12 and the distance index is 3. Next, according to the angle index and the distance index, determine the weight of the first point in the j-th sub-template with respect to the i-th prediction mode.
  • the above S102-A212 includes the following steps:
  • the weight of each point in the template is determined according to the angle index, the distance index, the size of the template and the size of the current block, and then the weight matrix formed by the weight of each point in the template is determined as the template weight.
  • the first parameter of this application is used to determine the weights.
  • the first parameter is also referred to as a weight index.
  • the offset and the first parameter may be determined in the following manner:
  • the template is only applied to the Y component, but it should be understood that the template can be applied to any component such as Y, Cb, Cr or any component of R, G, B, etc.
  • the selected first point is (x, y), and the weight derivation process of this first point with respect to the i-th prediction mode is as follows:
  • the input of this process includes: the width nCbW of the current block and the height nCbH of the current block, specifically as shown in Figure 19; the "division" angle index variable angleIdx of GPM; the distance index variable distanceIdx of GPM; the component index variable cIdx.
  • cIdx is 0, indicating the brightness component.
  • nW, nH, shift1, offset1, displacementX, displacementY, partFlip and shiftHor are derived as follows:
  • offsetY ((-nH)>>1)+(angleIdx ⁇ 16?(distanceIdx*nH)>>3:-((distanceIdx*nH)>>3))
  • offsetX ((-nW)>>1)+(angleIdx ⁇ 16?(distanceIdx*nW)>>3:-((distanceIdx*nW)>>3)
  • the weight wValue of the (x, y) position relative to the first prediction mode is derived as follows:
  • the first parameter weightIdx of the first point is determined according to the following formula:
  • weightIdx (((xL+offsetX) ⁇ 1)+1)*disLut[displacementX]+(((yL+offsetY) ⁇ 1)+1)*disLut[displacementY]
  • the weight of the first point (x, y) with respect to the i-th prediction mode is determined according to the weightIdx.
  • the ways of determining the weight of the first point with respect to the i-th prediction mode include but are not limited to the following:
  • the second parameter of the first point is determined; according to the second parameter of the first point, the weight of the first point with respect to the i-th prediction mode is determined.
  • the second parameter is also used to determine the weight.
  • the above-mentioned second parameter is also referred to as a weight index under the first component, and the first component may be a luma component, a chrominance component, and the like.
  • weightIdxL partFlip? 32+weightIdx:32-weightIdx
  • weightIdxL is 32–weightIdx.
  • weightIdxL is 32+weightIdx. It should be noted that 32 here is just a An example, the present application is not limited thereto.
  • the weight of the first point with respect to the i-th prediction mode is determined according to the first parameter of the first point, the first preset value, and the second preset value.
  • the weight of the first point with respect to the i-th prediction mode is limited to the first preset value or the second preset value, that is, the first point with respect to
  • the weight of the i-th prediction mode is either the first preset value or the second preset value, thereby reducing the complexity of calculating the weight of the first point with respect to the i-th prediction mode.
  • the present application does not limit specific values of the first preset value and the second preset value.
  • the first preset value is 1.
  • the second preset value is 0.
  • the weight of the first point with respect to the i-th prediction mode can be determined by the following formula:
  • wTemplateValue[x][y] is the weight of the first point (x, y), 1 in the above “1:0" is the first preset value, and 0 is the second preset value.
  • the j-th sub-template is corresponding to the i-th template middle.
  • the j-th sub-template is corresponding to the i-th template, and the i-th template is K A template of templates. For example, if the weight of the first point in the jth sub-template with respect to the first prediction mode is greater than the first preset value, then the jth sub-template is mapped to the first template. For another example, if the weight of the first point in the jth sub-template with respect to the first prediction mode is less than or equal to the first preset value, then the jth sub-template is mapped to the second template.
  • the present application does not limit the specific value of the above-mentioned first predicted value.
  • the above-mentioned first preset value is 0.
  • the above-mentioned first preset value is any positive number smaller than the median weight value. If the maximum weight value is 8, the median weight value is 4.
  • the weight of the first point with respect to the i-th prediction mode is greater than the first preset value, and the weight of the first point with respect to the i+1-th prediction mode is also greater than the first preset value, then, The j-th sub-template can be mapped to the i-th template, and the j-th sub-template can be mapped to the i+1-th template. At this time, the i-th template overlaps with the i+1-th template.
  • the first predicted value is 0 as an example, assuming that the jth sub-template is the sub-template 3 in Figure 18D, and the first point is the lower midpoint of the sub-template 3, the first point is determined according to the method above
  • the weight of a prediction mode is greater than 0, and the weight of the first point with respect to the second prediction mode is also greater than 0.
  • the sub-template 3 can be corresponding to the first template and the second template.
  • the above S102-A22 includes the following examples:
  • Example 1 if the weight of the first point with respect to the first prediction mode is greater than or equal to the second preset value, then the j-th sub-template is mapped to the first template.
  • the second default value is the median weight. If the maximum value of the weight is 8, the median weight is 4. If the weight of the first point of the jth sub-template with respect to the first prediction mode is greater than or equal to the median weight , then map the jth sub-template to the first template. Taking sub-template 2 in FIG. 18D as an example, according to the method above, it is determined that the weight of the first point of sub-template 2 with respect to the first prediction mode is 8, and this 8 is greater than the second prediction value (for example, 4), then the sub-template can be 2 corresponds to the first template.
  • the second prediction value for example, 4
  • Example 2 if the weight of the first point with respect to the first prediction mode is less than the second preset value, then the j-th sub-template is mapped to the second template.
  • the weight of the first point of the sub-template 4 with respect to the first prediction mode is determined to be 0 according to the above method, and this 0 is smaller than the second prediction value (for example, 4), then the sub-template can be 4 corresponds to the second template.
  • the specific implementation of determining K templates according to the weight derivation mode in case 1 is introduced.
  • the template of the current block is divided into K templates, Alternatively, divide the template of the current block into M sub-templates, and map the M sub-templates to K templates according to the weight derivation mode.
  • K templates in addition to determining K templates using the method of the above-mentioned case 1, K templates may also be determined according to the method of the following case 2.
  • the first correspondence includes different angle indexes or different weight derivation modes and K templates Correspondence between;
  • Fig. 20A and Fig. 20B show the weight matrix of GPM in 32x64 block and 64x32 block, and it can be seen that the intersection points of dividing lines and block boundaries are different under different shapes. Because the shape of the block changes but the angle of the dividing line does not change according to the shape of the block. For example, in the mode with index 52, there is an intersection point with the left boundary of the current block in the 32x64 block, but there is no intersection point with the left boundary of the current block in the 64x32 block, and the corresponding intersection point is at the lower boundary. That is to say, in a 32x64 block, the black part of pattern 52 is adjacent to the left template of the current block, while in a 64x32 block, the black part of pattern 52 has no adjacent part to the left template of the current block .
  • the embodiment of the present application sets different rules according to the length and width of the current block.
  • first correspondences are set for the three cases of length equal to width, length greater than width, and length less than width.
  • Each first correspondence can be the table shown in Table 5 above, including different angle indexes in this case.
  • the correspondence between different weight derivation modes and K templates can be the table shown in Table 5 above, including different angle indexes in this case.
  • a first corresponding relationship is set for each category, and the first corresponding relationship includes the category
  • the decoding end can determine the target first correspondence relationship corresponding to the current block from the first correspondence relationship corresponding to the preset different block sizes according to the size of the current block, such as the length and width of the current block, and according to In the weight derivation mode, K templates corresponding to the weight derivation mode are obtained from the first corresponding relationship of the target.
  • the first target correspondence includes correspondences between different angle indexes and K templates, it is necessary to determine the target angle index according to the weight derivation mode, and then obtain the target angle index from the first target correspondence according to the target angle index. Query the K templates corresponding to the target angle index in .
  • the decoder after determining K templates according to the above steps, performs the following step S103 to determine K prediction modes of the current block according to the K templates.
  • each of the K templates is used to determine a prediction mode, for example, the first prediction mode is determined using the first template among the K templates, and the second prediction mode is determined using the second template among the K templates. model.
  • the process of using each of the K templates to determine the corresponding prediction mode is the same, and the embodiment of the present application uses the i-th template in the K templates to determine the i-th prediction mode as an example for illustration.
  • the above S103 includes the following steps from S103-A1 to S103-A4:
  • the aforementioned at least one candidate prediction mode may be understood as a candidate prediction mode corresponding to the i-th prediction mode.
  • different prediction modes may correspond to different candidate prediction modes.
  • the candidate prediction modes corresponding to the two prediction modes may be the same.
  • the decoder when determining the i-th prediction mode, the decoder first judges whether the i-th prediction mode is determined through template matching.
  • a flag A is carried in the code stream, and the flag A is used to indicate whether the i-th prediction mode is determined through template matching. Exemplarily, if the value of the flag A is 1, it means that the i-th prediction mode is determined by template matching, and if the value of the flag A is 0, it means that the i-th prediction mode is not determined by template matching determined in a manner.
  • the decoding end decodes the code stream, obtains the flag A, and judges the value of the flag A. If the value of the flag A is 1, it is determined that the i-th prediction mode is determined by template matching. , the decoding end executes the method of the embodiment of the present application to obtain at least one candidate prediction mode, and determine the cost of the candidate prediction mode, and determine the jth prediction mode according to the cost of the candidate prediction mode.
  • both the encoding end and the decoding end default that the jth prediction mode is determined by template matching, so that when the decoding end determines the jth prediction mode, it uses template matching to determine the jth prediction mode by default.
  • j prediction modes then acquire at least one candidate prediction mode, and determine the cost of the candidate prediction mode, and determine the jth prediction mode according to the cost of the candidate prediction mode.
  • the above-mentioned jth prediction mode is an inter-frame prediction mode
  • the above-mentioned at least one candidate prediction mode includes one or more inter-frame prediction modes, such as skip, merge, normal inter-frame prediction mode, single At least one of directional forecasting, bidirectional forecasting, multi-hypothesis forecasting, and the like.
  • the above-mentioned jth prediction mode is an intra-frame prediction mode
  • the above-mentioned at least one candidate prediction mode includes at least one of DC (Direct Current, DC) mode, planar (PLANAR) mode, angle mode, etc.
  • the at least one candidate prediction mode includes an intra prediction mode in the MPM list.
  • At least one candidate prediction mode may also include IBC, palette and other modes.
  • the application does not limit the types of prediction modes and the number of prediction modes included in the at least one candidate prediction mode.
  • the above at least one candidate prediction mode is a preset mode.
  • the above at least one candidate prediction mode is a mode in the MPM list.
  • the above at least one candidate prediction mode is a set of candidate prediction modes determined according to some rules, such as equidistant screening.
  • the i-th template is predicted by using the candidate prediction mode, and the prediction value of the i-th template is determined.
  • the predicted value of the i-th template can be understood as a matrix composed of the predicted values of each pixel in the i-th template.
  • the cost of each candidate prediction mode is determined according to the prediction value of each candidate prediction mode with respect to the i-th template and the reconstruction value of the i-th template .
  • the loss of the candidate prediction mode for the i-th template is determined according to the prediction value of the candidate prediction mode for the i-th template and the reconstruction value of the i-th template, and the loss of the candidate prediction mode for the i-th template is determined. The cost of the candidate prediction mode.
  • the methods for determining the cost of candidate prediction modes in S103-A3 above include but are not limited to the following:
  • the first way is to determine the cost of the candidate prediction mode in the form of a matrix.
  • the loss samples are determined according to the prediction value of the i-th template and the reconstruction value of the i-th template in the candidate prediction mode, because the prediction value of the i-th template and the reconstruction value of the i-th template in the above-mentioned candidate prediction mode are both is a matrix, so the loss sample is also a matrix.
  • the absolute value of the difference between the predicted value of the candidate prediction mode with respect to the i-th template and the reconstructed value of the i-th template is determined as the loss sample.
  • determine the cost of the candidate prediction mode with respect to the i-th template for example, determine the sum of the losses of each point in the loss sample as the cost of the candidate prediction mode with respect to the i-th template.
  • the second way is to use point-by-point calculation to determine the cost of the candidate prediction mode, that is, the above S103-A3 includes the following steps:
  • S103-A323. Determine the cost of the candidate prediction mode according to the cost of the candidate prediction mode at each point in the i-th template.
  • the above i-th point can be understood as any point in the i-th template, that is to say, the process of determining the cost of each point in the i-th template is the same, just refer to the i-th point.
  • the candidate prediction mode is used to predict the i-th template, and the predicted value of the candidate prediction mode for the i-th template is obtained, and the corresponding predicted value of the i-th point in the predicted value of the i-th template is recorded as the i-th predictive value, record the reconstruction value corresponding to the i-th point in the reconstruction value of the i-th template as the i-th reconstruction value, and then determine the candidate prediction based on the i-th prediction value and the i-th reconstruction value.
  • the loss of the mode at the i-th point, and according to the loss of the candidate prediction mode at the i-th point determine the cost of the candidate prediction mode at the i-th point, for example, determine the loss of the candidate prediction mode at the i-th point is the
  • the cost of the candidate prediction mode at each point or multiple points in the i-th template is determined, and then the candidate prediction mode is determined according to the cost of each point or multiple points in the i-th template The cost of the i-th template.
  • the cost of the i-th template For example, the sum of the costs of the candidate prediction modes at each point in the i-th template is determined as the cost of the candidate prediction mode with respect to the i-th template, or the average cost of the candidate prediction modes at each point in the i-th template The value is determined as the cost of the candidate prediction mode with respect to the i-th template.
  • This application does not limit the determination of the cost of the candidate prediction mode with respect to the i-th template based on the cost of at least one point in the i-th template.
  • the cost of the candidate prediction mode at the i-th point (x, y) in the i-th template can be determined according to the following formula (1):
  • tempValueA[x][y] abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y]) (1)
  • the cost of the candidate prediction mode is determined according to the following formula (2):
  • abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y]) is the absolute value of the difference between the predicted value predTemplateSamplesCandA and the reconstructed value recTemplateSamples of the i-th template midpoint (x, y), and the difference
  • the absolute value is called the loss corresponding to the point (x, y).
  • tempValueA[x][y] can be considered as the cost of the candidate prediction mode at this point (x, y).
  • the total cost costCandA of the candidate prediction mode on the i-th template is the accumulation of the cost of each point on the i-th template.
  • SAD is used as an example to determine the cost of the candidate prediction mode.
  • the cost of the candidate prediction mode with respect to the i-th template can also be determined according to cost calculation methods such as SATD and MSE.
  • the cost of the candidate prediction mode with respect to the i-th template can be determined, and then the following steps of S103-A4 are performed.
  • the cost of the candidate prediction mode is determined through the above method, and the i-th prediction mode is determined according to the cost of the candidate prediction mode.
  • Example 1 The candidate prediction mode with the lowest cost among at least one candidate prediction mode is determined as the i-th prediction mode.
  • Example 2 Select one or more candidate prediction modes from at least one candidate prediction mode according to the cost of the candidate prediction modes; determine the jth prediction mode according to the one or more candidate prediction modes.
  • the decoding end selects a candidate prediction mode from one or more candidate prediction modes as the jth prediction mode.
  • the i-th prediction mode is determined from the above one or more candidate prediction modes according to an instruction from the coding end.
  • the above-mentioned one or more candidate prediction modes are M
  • the encoder sorts the M candidate prediction modes according to the cost, such as sorting the M candidate prediction modes according to the cost from small to large, or sorting the M candidate prediction modes according to the cost from large to small
  • the M candidate prediction modes are sorted, and a candidate prediction mode B is determined from the sorted M candidate prediction modes as the i-th prediction mode.
  • the coding end codes the identification of the candidate prediction mode B into the code stream.
  • the identification of the candidate prediction mode B may be the ranking number of the candidate prediction mode B among the M candidate prediction modes, or the candidate prediction mode B's schema index number. In this way, the decoding end obtains the identification of the candidate prediction mode B by decoding the code stream, and then according to the identification of the candidate prediction mode B, the candidate prediction mode corresponding to the identification of the candidate prediction mode B among the M candidate prediction modes determined above is determined. is the i-th prediction mode.
  • the decoding end obtains the alternative prediction mode of the current block; determines the cost of the alternative prediction mode when predicting the i-th template; The cost when the template is predicted and the cost of the one or more candidate prediction modes selected above with respect to the ith template, and a prediction mode is selected from the candidate prediction mode and the one or more candidate prediction modes as the ith template a forecasting model.
  • the above-mentioned candidate prediction modes of the current block include one or more of prediction modes of reconstructed decoded blocks surrounding the current block and/or preset prediction modes.
  • the preset prediction mode may include one or more of various modes such as DC mode, Bilinear mode, and Planar mode.
  • the decoding end obtains the candidate prediction modes of the current block, for example, takes one or more of the prediction modes of reconstructed decoded blocks around the current block and/or preset prediction modes as the candidate prediction modes of the current block.
  • determine the cost of each alternative prediction mode for predicting the template for example, use the alternative prediction mode to predict the current block to obtain a prediction value, compare the prediction value with the reconstruction value of the template, and obtain the alternative prediction
  • the cost of the mode where the cost of the alternative prediction mode can be the cost of SAD, SATD, etc.
  • a prediction mode is selected from the alternative prediction mode and the above one or more candidate prediction modes as the jth prediction mode, for example, the alternative The prediction mode with the lowest cost among the prediction mode and the above one or more candidate prediction modes is determined as the jth prediction mode.
  • the above-mentioned candidate prediction modes of the current block are different from the one or more candidate prediction modes determined above, that is, the decoding end uses the prediction modes and/or preset prediction Among the modes, the same prediction modes as those in the above one or more candidate prediction modes are deleted, and the remaining prediction modes are determined as the candidate prediction modes of the current block.
  • template matching can be "searched" on the basis of an initial motion information.
  • a prediction mode needs to determine a motion information. Some motion information can be determined within a certain range around an initial motion information, so as to determine some prediction modes. If an initial motion information is given, its motion vector is (xInit, yInit), set a search range such as a rectangular area from xInit-sR to xInit+sR in the horizontal direction, and from yInit-sR to yInit+sR in the vertical direction, where sR can be 2, 4, 8 etc.
  • Each motion vector in the rectangular area can be combined with other information of the initial motion information, such as a reference frame index and a prediction list flag, to determine a motion information, thereby determining a prediction mode.
  • the above at least one candidate prediction mode may include the determined prediction mode. For example, if GPM is used in the merge mode, if the template matching method is used to determine the first prediction mode, merge_gpm_idx0 can be used to determine an initial motion information from the mergeCandList. Then determine (2*sR+1)*(2*sR+1) pieces of motion information according to the above method, so as to determine some prediction modes, and these prediction modes are all merge modes, or called merge modes using template matching.
  • the process of determining the j-th prediction mode can also be further extended to a process of several layers from rough selection to fine selection.
  • the motion vector supports sub-pixel precision, such as 1/4, 1/8, 1/16 precision, etc.
  • the prediction mode with the least cost can be selected from the prediction modes containing the whole-pixel motion vector first, and then the cost can be further selected from the prediction mode and the prediction mode containing the sub-pixel motion vector whose motion vector is near the motion vector of this mode.
  • Minimal predictive mode For example, in the intra-frame prediction mode, according to the cost of the candidate prediction mode, one or several intra-frame prediction modes are selected at a certain granularity, and then the one or several intra-frame prediction modes and the finer-grained adjacent frames are selected. Then filter in the forecast mode.
  • the i-th prediction mode among the K prediction modes is determined by template matching, by obtaining at least one candidate prediction mode and using the candidate prediction mode to predict the template, the template in the candidate prediction mode is obtained The prediction value of the candidate prediction mode; according to the prediction value of the template in the candidate prediction mode and the reconstruction value of the template, the cost of the candidate prediction mode is obtained, and finally the jth prediction mode is obtained according to the cost of the candidate prediction mode.
  • the above embodiment is described by taking the determination process of the i-th prediction mode among the K prediction modes as an example.
  • the determination process of other prediction modes among the K prediction modes is consistent with the determination process of the i-th prediction mode.
  • K prediction modes can be determined according to the K templates, and then the K prediction modes are used to predict the current block to obtain the prediction value of the current block.
  • the K prediction modes are used to predict the current block to obtain the prediction value of the current block.
  • the weight is determined according to the weight derivation mode
  • K prediction values are determined according to K prediction modes
  • the K prediction values are weighted according to the weight
  • the weighted result is determined as the final prediction value.
  • the weight derivation mode is used to determine the weight when the prediction value of the current block is weighted.
  • the weight derivation mode may be a mode for deriving weights. For a block with a given length and width, each weight derivation mode can derive a weight matrix; for a block of the same size, the weight matrices derived from different weight derivation modes can be different.
  • the AWP of AVS3 has 56 weight export modes
  • the GPM of VVC has 64 weight export modes.
  • the above prediction process is performed in units of pixels, and the corresponding weights are also weights corresponding to pixels.
  • each of the K prediction modes is used to predict a certain pixel A in the current block, and K prediction values of the K prediction modes for the pixel A are obtained, according to The weight of pixel A weights the K predicted values to obtain the final predicted value of pixel A.
  • Performing the above steps for each pixel in the current block can obtain the final prediction value of each pixel in the current block, and the final prediction value of each pixel in the current block constitutes the final prediction value of the current block.
  • both the first prediction mode and the second prediction mode are intra-frame prediction modes
  • the first intra-frame prediction mode is used for prediction to obtain the first predicted value
  • the second The intra prediction mode performs prediction to obtain a second prediction value, and weights the first prediction value and the second prediction value according to the prediction weight to obtain a new prediction value.
  • the first intra-frame prediction mode is used to predict the pixel point A to obtain the first predicted value of the pixel point A
  • the second intra-frame prediction mode is used to predict the pixel point A to obtain the second predicted value of the pixel point A
  • the first prediction value and the second prediction value are weighted to obtain the final prediction value of the pixel point A.
  • the above-mentioned derivation mode based on the K prediction modes and weights, and determining the prediction value includes the following steps:
  • S104-AB25 Determine the predicted value according to the i-th predicted value, K-1 predicted values and weights.
  • the intra prediction mode is used for prediction
  • the first prediction value is obtained
  • the inter prediction mode is used for prediction , to obtain the second predicted value, and weight the first predicted value and the second predicted value according to the predicted weight to obtain a new predicted value.
  • the intra prediction mode is used to predict each point in the current block to obtain the predicted value of each point in the current block, and the predicted value of each point in the current block constitutes the first predicted value of the current block.
  • the inter-frame prediction mode determine a piece of motion information, determine the best matching block of the current block according to the motion information, and determine the best matching block as the second prediction value of the current block.
  • a point-by-point weighting operation is performed on the first prediction value and the second prediction value of the current block to obtain a new prediction value of the current block. For example, for pixel A in the current block, according to the prediction weight of pixel A, the first prediction value corresponding to pixel A in the first prediction value of the current block is compared with the pixel A in the second prediction value of the current block The corresponding second predicted value is weighted to obtain the final predicted value of pixel A.
  • the respective prediction weights of the K prediction modes can be determined according to the preset weight ratio, assuming that the prediction weight of the third prediction mode accounts for the entire 1/4 of the prediction weight, it can be determined that the prediction weight of the third prediction mode is 2, and the remaining 3/4 of the prediction weight is allocated to the first prediction mode and the second prediction mode.
  • the prediction weight 3 of the first prediction mode is derived according to the weight derivation mode, it is determined that the prediction weight of the first prediction mode is (3/4)*3, and the prediction weight of the second prediction mode is that of the first prediction mode The prediction weight is (3/4)*5.
  • the decoder before executing the method of the embodiment of the present application, the decoder needs to judge whether the current block is applicable to the template matching method, and if the decoder determines that the current block is applicable to the template matching method, then perform the above steps from S101 to S104, If the decoding end determines that the current block is not applicable to the template matching method, it uses other methods to determine the K prediction modes.
  • the decoder determines whether the current block is applicable to the template matching method through the following methods:
  • the decoder determines whether the current block is applicable to the template matching method according to the points included in the K templates.
  • the templates that can be obtained by the current block are on the left side and the top side of the current block, while the right side and the bottom side are not available, and such as The upper right and lower left are available in some cases and not in some cases.
  • a prediction mode cannot find the corresponding template or reconstructed adjacent regions.
  • the GPM indexes are 55, 56, and 57 in the weight matrix. The white area only exists in the lower right corner, and there is no template or adjacent reconstructed area directly adjacent to the white area.
  • the directly adjacent template or the adjacent reconstructed area can be found, but the adjacent area is very small, such as the index of GPM in the case of a square block is 59, the white in the weight matrix of 60 area.
  • the template directly adjacent to the current block or the reconstruction area directly adjacent is called an available area. If no available area is found or the available area is very small, the corresponding prediction mode is forcibly applied to the template matching or texture characteristic method. Not only will it not improve the compression efficiency, but it may be counterproductive. Because this prediction mode is different from the characteristics of the whole or most of the templates or adjacent reconstructed regions.
  • template matching or texture characteristics of adjacent reconstructed pixels are used for prediction modes with relatively large available templates, and template matching or texture characteristics of adjacent reconstructed pixels are not used for prediction modes with relatively small available templates .
  • the available templates corresponding to the first prediction mode are the white and gray areas in the template
  • the available templates corresponding to the second prediction mode are the black and gray areas in the template.
  • the first prediction mode If the area of the corresponding available template is larger, for example, greater than the preset value, the decoder determines that the first prediction mode is applicable to the template matching method. Similarly, it can be seen from FIG. 17B that the area of the available template corresponding to the second prediction mode is larger. For example, if the value is greater than the preset value, the decoder determines that the second prediction mode is also applicable to the template matching method.
  • the above S103 is performed to determine K prediction modes according to the K templates.
  • the above preset threshold may be 0.
  • the aforementioned preset threshold is a median weight, for example, 4.
  • the aforementioned preset threshold is a fixed value.
  • the aforementioned preset threshold is determined according to the size of the current block, for example, it is 1/m1 of the total points of the current block, and m1 is a positive number.
  • the preset threshold is determined according to the size of the template of the current block, for example, 1/m2 of the total number of points of the template of the current block, where m2 is a positive number.
  • K prediction modes are determined according to the weight derivation mode.
  • the decoder determines whether the current block is applicable according to the points included in the K templates after determining K templates based on the size of the current block and at least one of the weight derivation modes according to the above step S102 in the template matching method. Specifically, for the i-th template among the K templates, if the number of pixels included in the i-th template is greater than the preset threshold, it means that the i-th template used to determine the i-th prediction mode The available templates are large, and when the i-th template is used to determine the i-th prediction mode, the prediction effect can be improved.
  • the method determines the i-th prediction mode, not only will it not improve the compression efficiency, but it may have a negative effect.
  • Method 2 Decode the code stream at the end to obtain a first flag, which is used to indicate whether to use template matching to derive the prediction mode; and then determine whether the current block uses template matching to derive the prediction mode according to the first flag.
  • the first flag is used to indicate whether the current block uses the template matching method to derive the prediction mode; if the encoder determines that the current block uses the template matching method to derive the prediction mode , then set the first flag to 1, and write the first flag set to 1 into the code stream, if the encoder determines that the current block does not use template matching to derive the prediction mode, then set the first flag to 0 , and write the first flag set to 0 into the code stream.
  • the decoding end obtains the first flag by decoding the code stream, and determines whether the current block uses template matching to derive the prediction mode according to the first flag.
  • K prediction modes are determined according to the weight derivation mode.
  • the decoding end decodes the code stream to obtain the first flag.
  • the decoding end determines that the current block adopts the template matching method to derive the prediction mode, and then executes the above step of S102.
  • the weight derivation mode determine K prediction modes.
  • the decoding end decodes the code stream to obtain the first flag.
  • the decoding end determines that the current block does not use template matching to derive the prediction mode, and then determines K prediction modes in other ways, for example, by The weight derivation mode determines at least one of the K prediction modes of the current block.
  • the position where the weight changes constitutes a straight line (curve segment), or, as shown in FIG. 4 and FIG. 5 , the positions with the same weight in the transition region constitute a straight line (curve segment).
  • This straight line can be called a dividing line (or dividing line or dividing line).
  • the dividing line itself also has an angle. You can set the horizontal right angle to 0, and the angle increases counterclockwise. Then the dividing line may be horizontal 0 degrees, vertical 90 degrees, inclined such as 45 degrees, 135 degrees, and various other angles. If a block chooses to use a certain weight matrix, the corresponding texture is likely to show different characteristics on both sides of the dividing line.
  • one side of the dividing line is a texture with an angle.
  • the other side is a flatter texture. Since the dividing line itself also has an angle, it can be assumed that a point is obtained through angle prediction, which may be close to some textures of the current block, so this line is related to the two prediction modes of the current block .
  • a boundary in the horizontal direction matches a horizontal prediction mode, such as mode 18 in VVC; a boundary in the vertical direction matches a vertical intra prediction mode, such as mode 50 in VVC.
  • a boundary of 45 degrees can Match the intra prediction mode of 45 degrees from bottom left to top right, such as mode 66 in VVC, and also match the intra prediction mode of 225 degrees from top right to bottom left, such as mode 2 in VVC. Then the weight derivation mode can be matched to some intra prediction modes.
  • the weight derivation mode can also be an index of weight, for example, the 56 modes of AWP can be considered as 56 kinds of weight derivation modes, and the 64 modes of GPM of VVC can be considered as 64 kinds of weights export mode.
  • an intra-frame angle prediction mode corresponding to an angle close to the boundary line or an angle perpendicular to the boundary line is also high.
  • an intra-frame angle prediction mode corresponding to an angle close to the boundary line or an angle perpendicular to the boundary line is also high.
  • the GPM uses K different intra-frame prediction modes.
  • GPM needs to use one or a few intra-frame prediction modes. In this case A smaller range of intra-frame prediction modes can be provided for selection by the GPM, so as to save the overhead of selecting which intra-frame prediction mode is selected.
  • one predictor of GPM comes from intra prediction
  • one predictor comes from inter prediction.
  • the intra prediction mode used in this application is determined by the weight derivation mode by default.
  • the demarcation line of the weight derivation mode is in the horizontal direction, as shown in FIG. 4
  • the GPM indexes are 18, 19, 50, and 51
  • the intra prediction mode is determined to be the mode 18 in the horizontal direction.
  • the boundary line of the weight derivation mode is in the vertical direction, as shown in FIG. 4
  • the GPM index is 0, 1, 36, and 37
  • the intra prediction mode is determined to be the mode 50 in the vertical direction.
  • the type of the K prediction mode must first be determined.
  • the prediction mode is an intra prediction mode
  • the prediction can be determined according to the weight derivation mode model.
  • the method in the embodiment of the present application further includes:
  • Step 11-0 decoding the code stream to obtain a type flag, which is used to indicate whether the K prediction modes belong to the intra prediction mode;
  • Step 11-1 Determine the types of the K prediction modes according to the type flags.
  • mode0IsInter indicates the first prediction mode Whether the mode is an inter prediction mode
  • mode1IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode1IsInter is 1.
  • the value of the type flag when the value of the type flag is the second value, it indicates that the first prediction mode is an intra prediction mode, and the second prediction mode is an inter prediction mode. In this case, mode0IsInter is 0, and mode1IsInter is 1.
  • the value of the type flag when the value of the type flag is the third value, it indicates that the first prediction mode is an inter prediction mode, and the second prediction mode is an intra prediction mode. In this case, mode0IsInter is 1, and mode1IsInter is 0.
  • the value of the type flag is the fourth value, it indicates that both the first prediction mode and the second prediction mode are intra-frame prediction modes. In this case, mode0IsInter is 0, and mode1IsInter is 0.
  • the present application does not limit the specific values of the above-mentioned first value, second value, third value and fourth value.
  • the first value is 0.
  • the second value is 1.
  • the third value is 2.
  • the fourth value is 3.
  • the field intra_mode_idx can be used to indicate the type flag.
  • the type flag needs to be encoded into the code stream during encoding, and the decoder decodes the code stream to obtain the type flag, and The types of the first prediction mode and the second prediction mode are determined according to the type flag.
  • merge_gpm_partition_idx is the weight export mode or weight export index
  • intra_mode_idx is the type flag
  • merge_gpm_idx0 is the index value of the first motion information in the candidate list
  • merge_gpm_idx1 is the second motion information The index value in the candidate list.
  • the decoder determines the type of the K prediction mode according to the above type flag, if at least one of the K prediction modes is an intra prediction mode, the intra prediction mode is determined based on the weight derivation mode.
  • the intra prediction mode is determined based on the weight derivation mode.
  • the first prediction mode and the second prediction mode are both intra prediction modes, the first prediction mode and the second prediction mode are determined based on the weight derivation mode.
  • Two prediction models For another example, when one of the first prediction mode and the second prediction mode is an intra prediction mode, the intra prediction mode in the first prediction mode and the second prediction mode is determined based on the weight derivation mode.
  • the ways of determining at least one of the K prediction modes based on the weight derivation mode include but are not limited to the following:
  • Mode 1 if at least one of the K prediction modes is an intra prediction mode, then determine the angle index according to the weight derivation mode, and determine the intra prediction mode corresponding to the angle index as one of the K prediction modes .
  • angle index is used to indicate the boundary angle index of the weight.
  • the angle index is represented by the field angleIdx.
  • the above Table 2 shows the correspondence between merge_gpm_partition_idx and angleIdx.
  • the angle index can be derived according to the weight derivation mode.
  • the angle index has a corresponding relationship with the intra-frame prediction mode, that is, different angle indexes correspond to different intra-frame prediction modes.
  • the first prediction mode or the second prediction mode is an intra prediction mode
  • determine the angle index according to the weight derivation mode for example, according to the above Table 2
  • derive the angle index corresponding to the weight derivation mode angle index for example, according to the above Table 7
  • the intra prediction mode corresponding to the angle index is determined, for example, the angle index is 2, and the corresponding intra prediction mode is 42, and then the intra prediction mode 42 is determined as the first prediction mode or the second predictive mode.
  • Mode 2 if at least one of the K prediction modes is an intra prediction mode, then obtain the intra prediction mode corresponding to the weight derivation mode; determine at least one of the K prediction modes according to the intra prediction mode corresponding to the weight derivation mode one.
  • the first prediction mode and/or the second prediction mode are intra-frame prediction modes
  • the first prediction mode and/or the second prediction mode are frames corresponding to the weight derivation mode Determined in intra prediction mode.
  • the first prediction mode and/or the second prediction mode may be an intra prediction mode that is on the same straight line or approximately on the same straight line as the weight division line (also referred to as a boundary line).
  • the first prediction mode and/or the second prediction mode may be an intra-frame prediction mode perpendicular to or approximately perpendicular to the weight division line.
  • the dividing line of the weight is in the horizontal direction, as shown in Figure 4, the index of GPM is 18, 19, 50, 51, the first prediction mode and/or the second prediction mode is the mode 18 in the horizontal direction and the mode in the vertical direction Mode 50.
  • intra prediction modes corresponding to the weight derivation mode there are many types of intra prediction modes corresponding to the weight derivation mode, for example, including intra prediction modes parallel to the boundary line of weights, intra prediction modes perpendicular to the boundary line, and the like.
  • the present application may use a flag to indicate which one of the intra prediction modes corresponding to the weight derivation mode is specifically selected for the first prediction mode and/or the second prediction mode.
  • the first prediction mode is an intra prediction mode
  • use the second flag to indicate the correspondence between the first prediction mode and the intra prediction mode corresponding to the weight derivation mode, for example, the second The flag indicates that the first prediction mode is an intra prediction mode parallel to the boundary line of weights, or indicates that the first prediction mode is an intra prediction mode perpendicular to the boundary line of weights.
  • a third flag is used to indicate the correspondence between the second prediction mode and the intra prediction mode corresponding to the weight derivation mode, for example, the third flag indicates that the second prediction mode is An intra-frame prediction mode parallel to the dividing line of the weight, or indicating that the second prediction mode is an intra-frame prediction mode perpendicular to the dividing line of the weight.
  • the way of determining the first prediction mode and/or the second prediction mode according to the intra prediction mode corresponding to the weight derivation mode in the above method 2 includes but not limited to the following examples:
  • Example 1 if the first prediction mode is an intra prediction mode, obtain the second flag, and determine the intra prediction mode corresponding to the second flag in the intra prediction modes corresponding to the weight derivation mode as the first prediction mode.
  • Example 2 if the second prediction mode is the intra prediction mode, the third flag is obtained, and the intra prediction mode corresponding to the third flag among the intra prediction modes corresponding to the weight derivation mode is determined as the second prediction mode.
  • the intra prediction mode corresponding to the weight derivation mode includes at least one of an intra prediction mode parallel to the boundary line of the weight and an intra prediction mode perpendicular to the boundary line.
  • the second flag when the second flag is a fifth value, such as 0, it indicates that the first prediction mode is an intra prediction mode parallel to the boundary line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the second flag is the sixth value, such as 1, it indicates that the first prediction mode is an intra prediction mode perpendicular to the boundary line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the third flag when the third flag is a fifth value, such as 0, it indicates that the second prediction mode is an intra prediction mode parallel to the dividing line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the third flag is the sixth value, such as 1, it indicates that the second prediction mode is the intra prediction mode perpendicular to the boundary line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the intra prediction mode corresponding to the weight derivation mode includes at least one of an intra prediction mode parallel to the boundary line of the weight, an intra prediction mode perpendicular to the boundary line, and a planar mode.
  • the second flag when the second flag is a fifth value, such as 0, it indicates that the first prediction mode is an intra prediction mode parallel to the boundary line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the second flag is the sixth value, such as 1, it indicates that the first prediction mode is an intra prediction mode perpendicular to the boundary line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the second flag is the seventh value, such as 2, it indicates that the first prediction mode is planar mode.
  • the third flag when the third flag is a fifth value, such as 0, it indicates that the second prediction mode is an intra prediction mode parallel to the dividing line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the third flag is the sixth value, such as 1, it indicates that the second prediction mode is the intra prediction mode perpendicular to the boundary line of the weight among the intra prediction modes corresponding to the weight derivation mode.
  • the third flag is the seventh value, such as 2, it indicates that the second prediction mode is planar mode.
  • the field intra_gpm_idx0 is used to represent the second flag.
  • the field intra_gpm_idx1 is used to indicate the third flag.
  • the first prediction mode is an intra-frame prediction mode
  • the first prediction mode is determined according to the above-mentioned second flag
  • the second prediction mode is an intra-frame prediction mode
  • the first prediction mode is determined according to the above-mentioned third flag Two prediction models.
  • the second flag (intra_gpm_idx0) and/or the third flag (intra_gpm_idx1) are shown in Table 8.
  • the decoding end decodes the code stream shown in Table 8 to obtain the second flag and/or the third flag, and according to the second flag and/or the third flag, determine the first prediction mode and/or the second prediction mode, and then use The first prediction mode, the second prediction mode and the weight determine the prediction value.
  • the values of the second flag and the third flag are different.
  • a feasible way is to make the value of the second flag (intra_gpm_idx1) 0 and 1, and if intra_gpm_idx1 is greater than intra_gpm_idx0, add 1 to intra_gpm_idx1.
  • At least one of K prediction modes is determined according to the weight derivation mode, K preset values are determined according to the K prediction modes, and the K prediction values are weighted to obtain the final prediction value.
  • the decoding end determines the weight derivation mode of the current block by decoding the code stream; determines K templates according to at least one of the size of the current block and the weight derivation mode; and determines K according to the K templates. forecasting modes; according to the K forecasting modes and weight derived modes, the predicted value is determined. That is, the present application derives the mode based on the size and/or weight of the current block when determining K templates, so that the determined K templates are more in line with the actual situation, so that when using these K templates to determine the prediction mode, the prediction mode can be improved. The accuracy of the determination, and then use the accurately determined K prediction modes to achieve accurate prediction of the current block and improve the coding effect.
  • the prediction method of the present application is introduced above by taking the decoding end as an example, and the following description is made by taking the encoding end as an example.
  • FIG. 21 is a schematic flowchart of a prediction method provided by an embodiment of the present application, and the embodiment of the present application is applied to the video encoder shown in FIG. 1 and FIG. 2 .
  • the method of the embodiment of the present application includes:
  • the weight derivation mode is used to determine the weight used by the current block.
  • the weight derivation mode may be a mode for deriving weights.
  • each weight export mode can export a weight matrix; for blocks of the same size, different weight export modes export different weight matrices.
  • AWP has 56 weight derivation modes
  • GPM has 64 weight derivation modes.
  • the encoding end determines the weight derivation mode of the current block, including but not limited to the following:
  • the above weight export mode is the default mode, for example, the default weight export mode of the encoding end is the weight export mode with index number 44.
  • the second way is to determine the weight export mode according to the cost.
  • the encoder tries all possible combinations of K prediction modes and weight derivation modes, K is a positive integer greater than 1, selects the weight derivation mode in the combination with the smallest cost, and determines it as the weight derivation mode of the current block.
  • the above K prediction modes include the first prediction mode and the second prediction mode, assuming that there are 66 available prediction modes, and the first prediction mode has 66 possibilities. Since the second prediction mode is different from the first The prediction modes are different, so there are 65 second prediction modes, assuming that there are 63 weight derivation modes (taking GPM as an example), then this application may use any two different prediction modes and any weight derivation mode, a total of 66 ⁇ 65 ⁇ 63 possibilities. If the PCM prediction mode is not used. Then there are 65 ⁇ 64 ⁇ 63 possibilities. It can be seen that in the present application, the selectable prediction modes and the number of usable weight derivation modes can also be limited, and the number of combinations will be correspondingly reduced.
  • the encoding end may perform cost calculation on all possible combinations, and determine a combination with the smallest cost.
  • each combination includes K prediction modes and a weight derivation mode.
  • the encoder before determining the weight derivation mode of the current block, the encoder first needs to determine whether the current block uses K different prediction modes for weighted prediction processing. If the encoding end determines that the current block uses K different prediction modes for weighted prediction processing, it executes the above S201 to determine the weight derivation mode of the current block.
  • the encoding end may determine whether the current block uses K different prediction modes for weighted prediction processing by determining the prediction mode parameter of the current block.
  • the prediction mode parameter may indicate whether the current block can use GPM mode or AWP mode, that is, indicate whether the current block can use K different prediction modes for prediction processing.
  • the prediction mode parameter can be understood as a flag indicating whether the GPM mode or the AWP mode is used.
  • the encoding end may use a variable as the prediction mode parameter, so that the setting of the prediction mode parameter may be realized by setting the value of the variable.
  • the encoder can set the value of the prediction mode parameter to indicate that the current block uses GPM mode or AWP mode.
  • the encoder can set the variable The value of is set to 1.
  • the encoding end can set the value of the prediction mode parameter to indicate that the current block does not use GPM mode or AWP mode, specifically, the encoding end can Set the variable value to 0. Furthermore, in the embodiment of the present application, after the encoding end completes the setting of the prediction mode parameters, it can write the prediction mode parameters into the code stream and transmit it to the decoding end, so that the decoding end can analyze the code stream Get the prediction mode parameters.
  • the GPM mode or the AWP mode is a prediction method, specifically, K different prediction modes are determined for the current block, and then determined according to the K different prediction modes. K predicted values, and then the weights can be determined again, and the K predicted values are combined according to the weights, and finally a new predicted value can be obtained.
  • Fig. 15 is a schematic diagram of using two prediction modes to predict the current block. As shown in Fig. 15, when predicting the current block, the first prediction mode can be used to determine the first prediction value, while the second prediction mode can be used to determine The second predictive value can then use weights to combine the first predictive value and the second predictive value to finally obtain a new predictive value.
  • the size of the current block can be restricted.
  • the encoder can first determine the size parameter of the current block, and then determine whether the current block uses the GPM mode or the AWP mode according to the size parameter.
  • the limitation of the size of the block that can use the GPM mode or the AWP mode can also be realized through the limitation of the pixel parameters.
  • this application there may be a frame-level flag to determine whether the current frame to be encoded uses this application.
  • an intra frame such as an I frame
  • an inter frame such as a B frame, P frame
  • intra frames do not use this application
  • inter frames use this application.
  • Inter frames can also use intra prediction, thus inter frames are also likely to use this application.
  • a flag below the frame level and above the CU level such as tile, slice, patch, LCU, etc.
  • K is a positive integer greater than 1.
  • the present application does not limit the specific shape of the template of the current block.
  • the template of the current block includes at least one of an upper coded region and a left coded region of the current block.
  • the width of the upper encoded area is the same as the width of the current block, and the height of the left encoded area is the same as the height of the current block.
  • the division of the template is not fine enough, which leads to the problem of inaccurate determination of the prediction mode and large prediction error when the prediction mode is determined based on the imprecise template.
  • the embodiment of the present application implements fine division of templates by using at least one of the size of the current block and the weight derivation mode.
  • the process of determining K templates in S202 above based on the size and weight of the current block and at least one of the derived modes will be described in detail below in conjunction with the methods proposed in Case 1 and Case 2 below.
  • the embodiment of the present application can implement a finer division of templates through the weight derivation mode.
  • the above S202 includes the following steps:
  • the ways of dividing the template of the current block into K templates include but are not limited to the following:
  • Way 1 divide the template of the current block into K templates according to the boundary line of the weight matrix corresponding to the weight derivation mode.
  • the present application extends the boundary line of the weight matrix corresponding to the weight derivation mode of the current block to the template of the current block to divide the template.
  • the right side of the boundary line can be
  • the template is denoted as the first template
  • the template on the left side of the dividing line is denoted as the second template.
  • the first template corresponds to the first prediction mode
  • the second template corresponds to the second prediction mode.
  • the first template can be used to derive the first prediction mode
  • the second template can be used to derive the second prediction mode, thereby realizing the prediction mode. Accurately determine, improve the coding effect.
  • the first template and the second template divided according to the above method may not be rectangular.
  • the first template and the second template have hypotenuses, and the cost calculation for irregular templates is more complicated.
  • both the first template and the second template can be divided into rectangles.
  • the template of the current block is divided into K templates according to the dividing line of the weight matrix, which is simple and can realize accurate division of the templates.
  • the template of the current block may also be divided into K templates according to the second method as follows.
  • S202-A includes the following steps of S202-A1 and S202-A2:
  • the template of the current block is first divided into multiple sub-templates, for example, divided into M sub-templates, and then it is determined which template each sub-module corresponds to, and then the division of K templates is realized.
  • the embodiment of the present application does not limit the manner of dividing the foregoing sub-templates.
  • S202-A1 includes: dividing the template of the current block into M sub-templates according to the weight derivation mode.
  • Example 1 determine the weight matrix according to the weight derivation mode, extend the weight matrix to the template of the current block, for example, extend to the left and upward, and cover the weight matrix on the template of the current block.
  • the template of the current block includes the left area and the upper area of the current block, and the lower right rectangular area is the current block.
  • the weight matrix of the current block is extended to the template of the current block to cover the template of the current block, so that the template of the current block can be divided into M sub-templates according to the coverage of the template of the current block by the weight matrix.
  • the black template in FIG. 17D is divided into the first sub-template, the upper gray template is divided into the second sub-template, and the upper white template is divided into the second sub-template.
  • the black template on the left side in FIG. Divided into a fourth sub-template.
  • the present application does not limit the specific shapes of the above M sub-templates.
  • the above example 1 divides the M sub-templates into rectangles.
  • Example 2 according to the weight derivation mode, determine the boundary line of the weight, and extend the boundary line to the template of the current block, so as to divide the template of the current block into M sub-templates.
  • the boundary line of the weight is determined according to the weight derivation mode, the boundary line is extended to the template of the current block, and the upper template of the current block is divided into two parts. In this way, M sub-templates can be determined according to the templates divided by the weight dividing line.
  • the first template and the second template divided according to the above method may not be rectangular.
  • the dividing line is extended to the template of the current block to obtain The extension line of the dividing line in the template of the current block; use the extension line to divide the template of the current block into M rectangular sub-templates. For example, as shown in FIG. 17G , the first sub-template and the second sub-template are divided into rectangles using extension lines.
  • method 2 in addition to dividing the template of the current block into M sub-templates according to the above-mentioned weight derivation mode, the following implementation method 2 can also be used to divide the template of the current block into M sub-templates, as shown below.
  • the template of the current block is divided into M sub-templates, that is, the above S202-A1 includes the following steps:
  • both P and Q are integers less than or equal to M, and the sum of P and Q is equal to M.
  • the template of the current block includes several rows of pixels that have been coded above the current block and several columns of pixels that have been coded on the left side of the current block. Several pixel rows are recorded as the upper template of the current block, and several columns of pixels encoded on the left side of the current block are recorded as the left template of the current block.
  • the template of the current block also includes the coded area of the upper left corner of the current block, and/or includes the coded area of the lower left of the current block, etc.
  • the embodiment of the present application does not limit the specific template of the current block.
  • the division of the upper template and the left template among the templates of the current block is mainly described as an example.
  • implementation mode 2 there is no limit to the way of dividing the upper template of the current block into P sub-templates and/or dividing the left template of the current block into Q sub-templates, for example, it can be divided equally or according to a preset ratio. Divide, or divide according to the preset number of pixels, or divide according to the preset number of pixel rows or pixel columns, etc.
  • the manners of dividing the left template of the current block into P sub-templates in the above S202-A11 include but are not limited to the following:
  • Mode 1 divide the upper template into P sub-templates along the vertical direction.
  • the upper template of the current block is evenly divided into P equal parts along the vertical direction.
  • the upper template of the current block is divided into P sub-templates according to a preset ratio of sub-templates.
  • Mode 2 divide the upper template into P sub-templates according to the preset number of pixels.
  • the preset number of pixels is used as a minimum division unit, and the upper template of the current block is divided into P sub-templates.
  • the present application does not limit the specific arrangement manner of the preset pixel points.
  • n columns of pixels are used as a minimum division unit, and the upper template is divided into P sub-templates, where n is a positive integer.
  • the present application does not limit the specific value of the above n, for example, it is a preset value.
  • the length of the upper template of the current block is the same as the length of the current block, so that the aforementioned n can be determined according to the length of the current block, for example, the length of the current block is a positive integer multiple of n.
  • the length of the current block is 16, the n may be 2, 4, 8 and other values.
  • the division method of the left template may be the same as or different from the division method of the upper template of the current block.
  • the manners of dividing the left template of the current block into Q sub-templates in S202-A11 include but are not limited to the following:
  • Mode 1 divide the left template into Q sub-templates along the horizontal direction.
  • the left template of the current block is evenly divided into Q equal parts.
  • the left template of the current block is divided into Q sub-templates according to a preset ratio of sub-templates.
  • Mode 2 divide the left template into Q sub-templates according to the preset number of pixels.
  • the preset number of pixels is used as a minimum division unit, and the left template of the current block is divided into Q sub-templates.
  • m rows of pixels are used as a minimum division unit, and the left template is divided into Q sub-templates, where m is a positive integer.
  • the present application does not limit the specific value of the above m, for example, it is a preset value.
  • the width of the left template of the current block is the same as the width of the current block, so that the aforementioned m can be determined according to the width of the current block, for example, the width of the current block is a positive integer multiple of m.
  • the width of the current block is 16, then the m may be 2, 4, 8 and other values.
  • the above step S202-A2 is performed, that is, the M sub-templates are mapped to the K templates according to the weight derivation mode.
  • the template of the current block is divided into multiple sub-templates, for example, the template of the current block is divided into M sub-templates, and then, it is determined which sub-template each sub-template of the M sub-templates belongs to Templates, and then map M sub-templates to K templates to achieve fine and accurate division of templates.
  • mapping M sub-templates to K templates include but are not limited to the following:
  • Mode 1 according to the dividing line of the weight matrix, the M sub-templates are mapped to the K templates.
  • the sub-template can be mapped to the first template and the second template. At this time, the first template and the second template have overlapping part.
  • the sub-template is mapped to the first template or the second template by default.
  • the dividing line of the weight divides a sub-template into two parts. If the area of the sub-template in the first prediction mode is larger than the area in the second prediction mode, the sub-template is corresponding to the second prediction mode. in a template.
  • the above S202-A2 includes the following steps:
  • the weight of the pixel points in the sub-template it is determined which template the sub-template is divided into. For example, the weight of the pixel points in the sub-template is the same or basically the same as the weight corresponding to the first prediction mode, then The sub-template corresponds to the first template, and if the weight of the pixel in the sub-template is the same or substantially the same as the weight corresponding to the second prediction mode, then the sub-template is corresponding to the second template.
  • the embodiment of the present application uses the j-th sub-template among the M sub-templates as an example to illustrate the process of determining which template other sub-templates correspond to Just refer to the jth sub-template.
  • determining a pixel point in the jth sub-template, such as the weight of the first point, and according to the weight of the first point determine which pixel the j-th sub-template corresponds to in the template.
  • the above-mentioned first point is any point in the jth submodule.
  • the above-mentioned first point is a point on the boundary line between the jth sub-template and the current block.
  • the method of determining the weight of the first point in the j-th template with respect to each of the K prediction modes is the same, and the embodiment of the present application takes the determination of the weight of the first point with respect to the i-th prediction mode as an example for illustration.
  • the way of determining the weight of the first point in the j-th sub-template with respect to the i-th prediction mode in S202-A21 above includes but is not limited to the following examples:
  • the weight matrix of the current block is extended to the j-th sub-template, so that the weight matrix of the current block at least covers the first point in the j-th sub-template, and then the weight of the first point is obtained.
  • the weight of the first point in the j-th sub-template with respect to the i-th prediction mode is determined, that is, the above-mentioned S202-A21 includes the following steps:
  • the weight of the first point in the j-th sub-template with respect to the i-th prediction mode is derived through the weight derivation mode.
  • the angle index and the distance index are determined according to the weight derivation mode, where the angle index can be understood as The angle index of the dividing line of each weight exported by the weight export mode.
  • the angle index and distance index corresponding to the weight derivation mode can be determined according to the above Table 2. For example, if the weight derivation mode is 27, the corresponding angle index is 12 and the distance index is 3. Next, according to the angle index and the distance index, determine the weight of the first point in the j-th sub-template with respect to the i-th prediction mode.
  • the above S202-A212 includes the following steps:
  • S202-A212. Determine the first parameter of the first point according to the angle index, the distance index and the size of the current block;
  • the weight of each point in the template is determined according to the angle index, the distance index, the size of the template and the size of the current block, and then the weight matrix formed by the weight of each point in the template is determined as the template weight.
  • the first parameter of this application is used to determine the weights.
  • the first parameter is also referred to as a weight index.
  • the weight of the first point (x, y) with respect to the i-th prediction mode is determined according to the weightIdx.
  • the ways of determining the weight of the first point with respect to the i-th prediction mode include but are not limited to the following:
  • the second parameter of the first point is determined; according to the second parameter of the first point, the weight of the first point with respect to the i-th prediction mode is determined.
  • the second parameter is also used to determine the weight.
  • the above-mentioned second parameter is also referred to as a weight index under the first component, and the first component may be a luma component, a chrominance component, and the like.
  • weightIdxL partFlip? 32+weightIdx:32-weightIdx
  • wTemplateValue[x][y] is the weight of the first point (x, y) on the i-th prediction mode
  • weightIdxL is the second parameter of the first point (x, y)
  • weightIdxL is 32–weightIdx.
  • weightIdxL is 32+weightIdx. It should be noted that 32 here is just a An example, the present application is not limited thereto.
  • the weight of the first point with respect to the i-th prediction mode is determined according to the first parameter of the first point, the first preset value, and the second preset value.
  • the weight of the first point with respect to the i-th prediction mode is limited to the first preset value or the second preset value, that is, the first point with respect to
  • the weight of the i-th prediction mode is either the first preset value or the second preset value, thereby reducing the complexity of calculating the weight of the first point with respect to the i-th prediction mode.
  • the present application does not limit specific values of the first preset value and the second preset value.
  • the first preset value is 1.
  • the second preset value is 0.
  • the weight of the first point with respect to the i-th prediction mode can be determined by the following formula:
  • wTemplateValue[x][y] is the weight of the first point (x, y), 1 in the above “1:0" is the first preset value, and 0 is the second preset value.
  • the j-th sub-template is corresponding to the i-th template middle.
  • the j-th sub-template is corresponding to the i-th template, and the i-th template is K A template of templates. For example, if the weight of the first point in the jth sub-template with respect to the first prediction mode is greater than the first preset value, then the jth sub-template is mapped to the first template. For another example, if the weight of the first point in the jth sub-template with respect to the first prediction mode is less than or equal to the first preset value, then the jth sub-template is mapped to the second template.
  • the present application does not limit the specific value of the above-mentioned first predicted value.
  • the above-mentioned first preset value is 0.
  • the above-mentioned first preset value is any positive number smaller than the median weight value. If the maximum weight value is 8, the median weight value is 4.
  • the weight of the first point with respect to the i-th prediction mode is greater than the first preset value, and the weight of the first point with respect to the i+1-th prediction mode is also greater than the first preset value, then,
  • the j-th sub-template can be mapped to the i-th template, and the j-th sub-template can be mapped to the i+1-th template.
  • the i-th template overlaps with the i+1-th template.
  • the first point is the midpoint of the lower side of the sub-template 3, and the first point is determined according to the method above.
  • the weight of a prediction mode is greater than 0, and the weight of the first point with respect to the second prediction mode is also greater than 0.
  • the sub-template 3 can be corresponding to the first template and the second template.
  • the above S202-A22 includes the following examples:
  • Example 1 if the weight of the first point with respect to the first prediction mode is greater than or equal to the second preset value, then the j-th sub-template is mapped to the first template.
  • Example 2 if the weight of the first point with respect to the first prediction mode is less than the second preset value, then the j-th sub-template is mapped to the second template.
  • the specific implementation of determining K templates according to the weight derivation mode in case 1 is introduced.
  • the template of the current block is divided into K templates, Alternatively, divide the template of the current block into M sub-templates, and map the M sub-templates to K templates according to the weight derivation mode.
  • K templates in addition to determining K templates using the method of the above-mentioned case 1, K templates may also be determined according to the method of the following case 2.
  • the first correspondence includes different angle indexes or different weight derivation modes and K templates Correspondence between;
  • Fig. 10A and Fig. 10B show the weight matrix of GPM in 32x64 block and 64x32 block, and it can be seen that the intersection points of dividing lines and block boundaries in different shapes are not the same. Because the shape of the block changes but the angle of the dividing line does not change according to the shape of the block. For example, in the mode with index 52, there is an intersection point with the left boundary of the current block in the 32x64 block, but there is no intersection point with the left boundary of the current block in the 64x32 block, and the corresponding intersection point is at the lower boundary. That is to say, in a 32x64 block, the black part of pattern 52 is adjacent to the left template of the current block, while in a 64x32 block, the black part of pattern 52 has no adjacent part to the left template of the current block .
  • the embodiment of the present application sets different rules according to the length and width of the current block.
  • first correspondences are set for the three cases of length equal to width, length greater than width, and length less than width.
  • Each first correspondence can be the table shown in Table 5 above, including different angle indexes in this case.
  • the correspondence between different weight derivation modes and K templates can be the table shown in Table 5 above, including different angle indexes in this case.
  • a first corresponding relationship is set for each category, and the first corresponding relationship includes the category
  • the encoding end can determine the target first correspondence relationship corresponding to the current block from the first correspondence relationship corresponding to the preset different block sizes according to the size of the current block, such as the length and width of the current block, and according to In the weight derivation mode, K templates corresponding to the weight derivation mode are obtained from the first corresponding relationship of the target.
  • the first target correspondence includes correspondences between different angle indexes and K templates, it is necessary to determine the target angle index according to the weight derivation mode, and then obtain the target angle index from the first target correspondence according to the target angle index. Query the K templates corresponding to the target angle index in .
  • step S203 is performed to determine K prediction modes of the current block according to the K templates.
  • the above S203 includes the following steps from S203-A1 to S203-A4:
  • the aforementioned at least one candidate prediction mode may be understood as a candidate prediction mode corresponding to the i-th prediction mode.
  • different prediction modes may correspond to different candidate prediction modes.
  • the candidate prediction modes corresponding to the two prediction modes may be the same.
  • the encoder when determining the i-th prediction mode, the encoder first judges whether the i-th prediction mode is determined by template matching.
  • a flag A is acquired, and the flag A is used to indicate whether the i-th prediction mode is determined through template matching.
  • the encoding end judges the value of the flag A, and if the value of the flag A is 1, it is determined that the i-th prediction mode is determined by template matching, and at this time, the encoding end executes the embodiment of the present application The method of obtaining at least one candidate prediction mode, and determining the cost of the candidate prediction mode, and determining the jth prediction mode according to the cost of the candidate prediction mode.
  • the encoding end defaults that the i-th prediction mode is determined by template matching, so that when determining the i-th prediction mode, the encoding end uses template matching to determine the i-th prediction mode by default , and then acquire at least one candidate prediction mode, and determine the cost of the candidate prediction mode, and determine the jth prediction mode according to the cost of the candidate prediction mode.
  • the above-mentioned at least one candidate prediction mode includes one or more inter-frame prediction modes, such as skip, merge, normal inter-frame prediction mode, single At least one of directional forecasting, bidirectional forecasting, multi-hypothesis forecasting, and the like.
  • the above-mentioned jth prediction mode is an intra-frame prediction mode
  • the above-mentioned at least one candidate prediction mode includes at least one of DC (Direct Current, DC) mode, planar (PLANAR) mode, angle mode, etc.
  • the at least one candidate prediction mode includes an intra prediction mode in the MPM list.
  • At least one candidate prediction mode may also include IBC, palette and other modes.
  • the application does not limit the types of prediction modes and the number of prediction modes included in the at least one candidate prediction mode.
  • the above at least one candidate prediction mode is a preset mode.
  • the above at least one candidate prediction mode is a mode in the MPM list.
  • the above at least one candidate prediction mode is a set of candidate prediction modes determined according to some rules, such as equidistant screening.
  • the i-th template is predicted by using the candidate prediction mode, and the prediction value of the i-th template is determined.
  • the predicted value of the i-th template can be understood as a matrix composed of the predicted values of each pixel in the i-th template.
  • the cost of each candidate prediction mode is determined according to the prediction value of each candidate prediction mode with respect to the i-th template and the reconstruction value of the i-th template .
  • the loss of the candidate prediction mode for the i-th template is determined according to the prediction value of the candidate prediction mode for the i-th template and the reconstruction value of the i-th template, and the loss of the candidate prediction mode for the i-th template is determined. The cost of the candidate prediction mode.
  • the methods for determining the cost of the candidate prediction mode in S203-A3 above include but are not limited to the following:
  • the first way is to determine the cost of the candidate prediction mode in the form of a matrix.
  • the second way is to use point-by-point calculation to determine the cost of the candidate prediction mode, that is, the above S203-A3 includes the following steps:
  • S203-A323. Determine the cost of the candidate prediction mode according to the cost of the candidate prediction mode at each point in the i-th template.
  • the above i-th point can be understood as any point in the i-th template, that is to say, the process of determining the cost of each point in the i-th template is the same, just refer to the i-th point.
  • the candidate prediction mode is used to predict the i-th template, and the predicted value of the candidate prediction mode for the i-th template is obtained, and the corresponding predicted value of the i-th point in the predicted value of the i-th template is recorded as the i-th predictive value, record the reconstruction value corresponding to the i-th point in the reconstruction value of the i-th template as the i-th reconstruction value, and then determine the candidate prediction based on the i-th prediction value and the i-th reconstruction value.
  • the loss of the mode at the i-th point, and according to the loss of the candidate prediction mode at the i-th point determine the cost of the candidate prediction mode at the i-th point, for example, determine the loss of the candidate prediction mode at the i-th point is the
  • the cost of the candidate prediction mode at each point or multiple points in the i-th template is determined, and then the candidate prediction mode is determined according to the cost of each point or multiple points in the i-th template The cost of the i-th template.
  • the cost of the i-th template For example, the sum of the costs of the candidate prediction modes at each point in the i-th template is determined as the cost of the candidate prediction mode with respect to the i-th template, or the average cost of the candidate prediction modes at each point in the i-th template The value is determined as the cost of the candidate prediction mode with respect to the i-th template.
  • This application does not limit the determination of the cost of the candidate prediction mode with respect to the i-th template based on the cost of at least one point in the i-th template.
  • the cost of the candidate prediction mode at the i-th point (x, y) in the i-th template can be determined according to the following formula (3):
  • tempValueA[x][y] abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y])(3)
  • the cost of the candidate prediction mode is determined according to the following formula (4):
  • abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y]) is the absolute value of the difference between the predicted value predTemplateSamplesCandA and the reconstructed value recTemplateSamples of the i-th template midpoint (x, y), and the difference
  • the absolute value is called the loss corresponding to the point (x, y).
  • tempValueA[x][y] can be considered as the cost of the candidate prediction mode at this point (x, y).
  • the total cost costCandA of the candidate prediction mode on the i-th template is the accumulation of the cost of each point on the i-th template.
  • SAD is used as an example to determine the cost of the candidate prediction mode.
  • the cost of the candidate prediction mode with respect to the i-th template can also be determined according to cost calculation methods such as SATD and MSE.
  • the cost of the candidate prediction mode with respect to the i-th template can be determined, and then the following steps of S203-A4 are performed.
  • S203-A Determine the i-th prediction mode according to the cost of at least one candidate prediction mode.
  • the cost of the candidate prediction modes is determined through the above method, and the i-th prediction mode is determined according to the costs of each candidate prediction mode.
  • Example 1 The candidate prediction mode with the lowest cost among at least one candidate prediction mode is determined as the i-th prediction mode.
  • Example 2 Select one or more candidate prediction modes from at least one candidate prediction mode according to the cost of the candidate prediction modes; determine the jth prediction mode according to the one or more candidate prediction modes.
  • the coding end selects a candidate prediction mode from one or more candidate prediction modes as the jth prediction mode.
  • the above-mentioned one or more candidate prediction modes are M
  • the encoder sorts the M candidate prediction modes according to the cost, such as sorting the M candidate prediction modes according to the cost from small to large, or sorting the M candidate prediction modes according to the cost from large to small
  • the M candidate prediction modes are sorted, and a candidate prediction mode B is determined from the sorted M candidate prediction modes as the i-th prediction mode.
  • the coding end codes the identification of the candidate prediction mode B into the code stream.
  • the identification of the candidate prediction mode B may be the ranking number of the candidate prediction mode B among the M candidate prediction modes, or the candidate prediction mode B's schema index number.
  • the decoding end obtains the identification of the candidate prediction mode B by decoding the code stream, and then according to the identification of the candidate prediction mode B, the candidate prediction mode corresponding to the identification of the candidate prediction mode B among the M candidate prediction modes determined above is determined. is the i-th prediction mode.
  • the encoder obtains the alternative prediction mode of the current block; determines the cost of the alternative prediction mode for predicting the i-th template; The cost when the template is predicted and the cost of the one or more candidate prediction modes selected above with respect to the ith template, and a prediction mode is selected from the candidate prediction mode and the one or more candidate prediction modes as the ith template a forecasting model.
  • the above-mentioned candidate prediction modes of the current block include one or more of prediction modes of reconstructed coded blocks around the current block and/or preset prediction modes.
  • the preset prediction mode may include one or more of various modes such as DC mode, Bilinear mode, and Planar mode.
  • the decoding end obtains the candidate prediction modes of the current block, for example, takes one or more of the prediction modes of reconstructed decoded blocks around the current block and/or preset prediction modes as the candidate prediction modes of the current block.
  • determine the cost of each alternative prediction mode for predicting the template for example, use the alternative prediction mode to predict the current block to obtain a prediction value, compare the prediction value with the reconstruction value of the template, and obtain the alternative prediction
  • the cost of the mode where the cost of the alternative prediction mode can be the cost of SAD, SATD, etc.
  • a prediction mode is selected from the alternative prediction mode and the above one or more candidate prediction modes as the jth prediction mode, for example, the alternative The prediction mode with the lowest cost among the prediction mode and the above one or more candidate prediction modes is determined as the jth prediction mode.
  • the above-mentioned candidate prediction modes of the current block are different from the one or more candidate prediction modes determined above, that is, the decoding end uses the prediction modes and/or preset prediction Among the modes, the same prediction modes as those in the above one or more candidate prediction modes are deleted, and the remaining prediction modes are determined as the candidate prediction modes of the current block.
  • template matching can be "searched" on the basis of an initial motion information.
  • a prediction mode needs to determine a motion information. Some motion information can be determined within a certain range around an initial motion information, so as to determine some prediction modes. If an initial motion information is given, its motion vector is (xInit, yInit), set a search range such as a rectangular area from xInit-sR to xInit+sR in the horizontal direction, and from yInit-sR to yInit+sR in the vertical direction, where sR can be 2, 4, 8 etc.
  • Each motion vector in the rectangular area can be combined with other information of the initial motion information, such as a reference frame index and a prediction list flag, to determine a motion information, thereby determining a prediction mode.
  • the above at least one candidate prediction mode may include the determined prediction mode. For example, if GPM is used in the merge mode, if the template matching method is used to determine the first prediction mode, merge_gpm_idx0 can be used to determine an initial motion information from the mergeCandList. Then determine (2*sR+1)*(2*sR+1) pieces of motion information according to the above method, so as to determine some prediction modes, and these prediction modes are all merge modes, or called merge modes using template matching.
  • the process of determining the j-th prediction mode can also be further extended to a process of several layers from rough selection to fine selection.
  • the motion vector supports sub-pixel precision, such as 1/4, 1/8, 1/16 precision, etc.
  • the prediction mode with the least cost can be selected from the prediction modes containing the whole-pixel motion vector first, and then the cost can be further selected from the prediction mode and the prediction mode containing the sub-pixel motion vector whose motion vector is near the motion vector of this mode.
  • Minimal predictive mode For example, in the intra-frame prediction mode, according to the cost of the candidate prediction mode, one or several intra-frame prediction modes are selected at a certain granularity, and then the one or several intra-frame prediction modes and the finer-grained adjacent frames are selected. Then filter in the forecast mode.
  • the i-th prediction mode among the K prediction modes is determined by template matching, by obtaining at least one candidate prediction mode and using the candidate prediction mode to predict the template, the template in the candidate prediction mode is obtained The prediction value of the candidate prediction mode; according to the prediction value of the template in the candidate prediction mode and the reconstruction value of the template, the cost of the candidate prediction mode is obtained, and finally the jth prediction mode is obtained according to the cost of the candidate prediction mode.
  • the above embodiment is described by taking the determination process of the i-th prediction mode among the K prediction modes as an example.
  • the determination process of other prediction modes among the K prediction modes is consistent with the determination process of the i-th prediction mode.
  • K prediction modes can be determined according to the K templates, and then the K prediction modes are used to predict the current block to obtain the prediction value of the current block. For details, refer to the description in S204 below.
  • the weight is determined according to the weight derivation mode
  • K prediction values are determined according to K prediction modes
  • the K prediction values are weighted according to the weight
  • the weighted result is determined as the final prediction value.
  • the weight derivation mode is used to determine the weight when the prediction value of the current block is weighted.
  • the weight derivation mode may be a mode for deriving weights. For a block with a given length and width, each weight derivation mode can derive a weight matrix; for a block of the same size, the weight matrices derived from different weight derivation modes can be different.
  • the AWP of AVS3 has 56 weight export modes
  • the GPM of VVC has 64 weight export modes.
  • the encoder when determining the prediction value based on the K prediction modes and weights, the encoder can first determine the corresponding The predicted value is weighted to the predicted value corresponding to each prediction mode to obtain the final predicted value.
  • the above prediction process is performed in units of pixels, and the corresponding weights are also weights corresponding to pixels.
  • each of the K prediction modes is used to predict a certain pixel A in the current block, and K prediction values of the K prediction modes for the pixel A are obtained, according to The weight of pixel A weights the K predicted values to obtain the final predicted value of pixel A.
  • Performing the above steps for each pixel in the current block can obtain the final prediction value of each pixel in the current block, and the final prediction value of each pixel in the current block constitutes the final prediction value of the current block.
  • both the first prediction mode and the second prediction mode are intra-frame prediction modes
  • the first intra-frame prediction mode is used for prediction to obtain the first predicted value
  • the second The intra prediction mode performs prediction to obtain a second prediction value, and weights the first prediction value and the second prediction value according to the prediction weight to obtain a new prediction value.
  • the first intra-frame prediction mode is used to predict the pixel point A to obtain the first predicted value of the pixel point A
  • the second intra-frame prediction mode is used to predict the pixel point A to obtain the second predicted value of the pixel point A
  • the first prediction value and the second prediction value are weighted to obtain the final prediction value of the pixel point A.
  • the above-mentioned derivation mode based on the K prediction modes and weights, and determining the prediction value includes the following steps:
  • S204-AB25 Determine the predicted value according to the i-th predicted value, K-1 predicted values and weights.
  • the intra prediction mode is used for prediction
  • the first prediction value is obtained
  • the inter prediction mode is used for prediction , to obtain the second predicted value, and weight the first predicted value and the second predicted value according to the predicted weight to obtain a new predicted value.
  • the intra prediction mode is used to predict each point in the current block to obtain the predicted value of each point in the current block, and the predicted value of each point in the current block constitutes the first predicted value of the current block.
  • the inter-frame prediction mode determine a piece of motion information, determine the best matching block of the current block according to the motion information, and determine the best matching block as the second prediction value of the current block.
  • a point-by-point weighting operation is performed on the first prediction value and the second prediction value of the current block to obtain a new prediction value of the current block. For example, for pixel A in the current block, according to the prediction weight of pixel A, the first prediction value corresponding to pixel A in the first prediction value of the current block is compared with the pixel A in the second prediction value of the current block The corresponding second predicted value is weighted to obtain the final predicted value of pixel A.
  • the encoding end before executing the method of the embodiment of the present application, the encoding end needs to judge whether the current block is applicable to the template matching method, and if the encoding end determines that the current block is applicable to the template matching method, then perform the above steps from S201 to S204, If the encoding end determines that the current block is not applicable to the template matching method, it uses other methods to determine the K prediction modes.
  • the encoder determines whether the current block is applicable to the template matching method according to the points included in the K templates.
  • the prediction mode with relatively large available templates uses template matching or the texture characteristics of adjacent reconstructed pixels, while the prediction mode with relatively small available templates does not use template matching or the texture characteristics of adjacent reconstructed pixels.
  • the above S203 is performed to determine K prediction modes according to the K templates.
  • the above preset threshold may be 0.
  • the aforementioned preset threshold is a median weight, for example, 4.
  • the aforementioned preset threshold is a fixed value.
  • the aforementioned preset threshold is determined according to the size of the current block, for example, it is 1/m1 of the total points of the current block, and m1 is a positive number.
  • the preset threshold is determined according to the size of the template of the current block, for example, 1/m2 of the total number of points of the template of the current block, where m2 is a positive number.
  • K prediction modes are determined according to the weight derivation mode.
  • the encoding end determines whether the current block is applicable to the template according to the points included in the K templates after determining K templates based on at least one of the current block size and weight derivation mode according to the above-mentioned step S202 matching method. Specifically, for the i-th template among the K templates, if the number of pixels included in the i-th template is greater than the preset threshold, it means that the i-th template used to determine the i-th prediction mode The available templates are large, and when the i-th template is used to determine the i-th prediction mode, the prediction effect can be improved.
  • the method determines the i-th prediction mode, not only will it not improve the compression efficiency, but it may have a negative effect.
  • the encoder writes a first flag into the code stream, and the first flag is used to indicate whether the current block uses template matching to derive the prediction mode. If the encoder determines that the current block uses template matching to derive the prediction mode, Then set the first flag to 1, and write the first flag set to 1 into the code stream, and if the encoder determines that the current block does not use template matching to derive the prediction mode, then set the first flag to 0, And write the first flag set to 0 into the code stream. In this way, after obtaining the code stream, the decoding end obtains the first flag by decoding the code stream, and determines whether the current block uses template matching to derive the prediction mode according to the first flag.
  • At least one of the K prediction modes is determined according to the weight derivation mode.
  • the position where the weight changes constitutes a straight line (curve segment), or, as shown in FIG. 4 and FIG. 5 , the positions with the same weight in the transition region constitute a straight line (curve segment).
  • This straight line can be called a dividing line (or dividing line or dividing line).
  • the dividing line itself also has an angle. You can set the horizontal right angle to 0, and the angle increases counterclockwise. Then the dividing line may be horizontal 0 degrees, vertical 90 degrees, inclined such as 45 degrees, 135 degrees, and various other angles. If a block chooses to use a certain weight matrix, the corresponding texture is likely to show different characteristics on both sides of the dividing line.
  • one side of the dividing line is a texture with an angle.
  • the other side is a flatter texture. Since the dividing line itself also has an angle, it can be assumed that a point is obtained through angle prediction, which may be close to some textures of the current block, so this line is related to the two prediction modes of the current block .
  • the boundary line is obtained from a point through angle prediction, then at least one angle prediction mode can be found, and this angle prediction mode can approximate the boundary line.
  • the weight derivation mode can also be an index of weight, for example, the 56 modes of AWP can be considered as 56 kinds of weight derivation modes, and the 64 modes of GPM of VVC can be considered as 64 kinds of weights export mode.
  • an intra-frame angle prediction mode corresponding to an angle close to the boundary line or an angle perpendicular to the boundary line is also high.
  • an intra-frame angle prediction mode corresponding to an angle close to the boundary line or an angle perpendicular to the boundary line is also high.
  • one predictor of GPM comes from intra prediction
  • one predictor comes from inter prediction.
  • the intra prediction mode used in this application is determined by the weight derivation mode by default.
  • the demarcation line of the weight derivation mode is in the horizontal direction, as shown in FIG. 4
  • the GPM indexes are 18, 19, 50, and 51
  • the intra prediction mode is determined to be the mode 18 in the horizontal direction.
  • the boundary line of the weight derivation mode is in the vertical direction, as shown in FIG. 4
  • the GPM index is 0, 1, 36, and 37
  • the intra prediction mode is determined to be the mode 50 in the vertical direction.
  • the type of the K prediction mode must first be determined.
  • the prediction mode is an intra prediction mode
  • the prediction can be determined according to the weight derivation mode model.
  • the method in the embodiment of the present application further includes:
  • Step 21-0 acquiring a type flag, which is used to indicate whether the K prediction modes belong to the intra prediction mode
  • Step 21-1 Determine the types of the K prediction modes according to the type flags.
  • mode0IsInter indicates whether the first prediction mode is an inter prediction mode
  • mode1IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter indicates whether the second prediction mode is an inter prediction mode
  • mode0IsInter is 1
  • mode1IsInter indicates whether the second prediction mode is an inter prediction mode
  • the value of the type flag when the value of the type flag is the second value, it indicates that the first prediction mode is an intra prediction mode, and the second prediction mode is an inter prediction mode. In this case, mode0IsInter is 0, and mode1IsInter is 1.
  • the value of the type flag when the value of the type flag is the third value, it indicates that the first prediction mode is an inter prediction mode, and the second prediction mode is an intra prediction mode. In this case, mode0IsInter is 1, and mode1IsInter is 0.
  • the value of the type flag is the fourth value, it indicates that both the first prediction mode and the second prediction mode are intra-frame prediction modes. At this time, mode0IsInter is 0, and mode1IsInter is 0.
  • the present application does not limit the specific values of the above-mentioned first value, second value, third value and fourth value.
  • the first value is 0.
  • the second value is 1.
  • the third value is 2.
  • the fourth value is 3.
  • the field intra_mode_idx can be used to indicate the type flag.
  • the encoder After the encoder determines the types of the first prediction mode and the second prediction mode according to the type flag, it needs to encode the type flag into the code stream during encoding, so that the first prediction mode and the second prediction mode can be determined according to the type flag.
  • Two types of prediction models Two types of prediction models.
  • the encoder determines the type of the K prediction mode according to the above type flag, if at least one of the K prediction modes is an intra prediction mode, the intra prediction mode is determined based on the weight derivation mode.
  • the intra prediction mode is determined based on the weight derivation mode.
  • the first prediction mode and the second prediction mode are both intra prediction modes, the first prediction mode and the second prediction mode are determined based on the weight derivation mode.
  • Two prediction models For another example, when one of the first prediction mode and the second prediction mode is an intra prediction mode, the intra prediction mode in the first prediction mode and the second prediction mode is determined based on the weight derivation mode.
  • the ways of determining at least one of the K prediction modes based on the weight derivation mode include but are not limited to the following:
  • Mode 1 if at least one of the K prediction modes is an intra prediction mode, then determine the angle index according to the weight derivation mode, and determine the intra prediction mode corresponding to the angle index as one of the K prediction modes .
  • the angle index is represented by the field angleIdx.
  • the above Table 2 shows the correspondence between merge_gpm_partition_idx and angleIdx.
  • the angle index can be derived according to the weight derivation mode.
  • the angle index has a corresponding relationship with the intra-frame prediction mode, that is, different angle indexes correspond to different intra-frame prediction modes.
  • the angle index is determined according to the weight derivation mode.
  • the intra prediction mode corresponding to the angle index is determined, for example, the angle index is 2, and the corresponding intra prediction mode is 42, and then the intra prediction mode 42 is determined as the first prediction mode or the second prediction mode.
  • Mode 2 if at least one of the K prediction modes is an intra prediction mode, then obtain the intra prediction mode corresponding to the weight derivation mode; determine at least one of the K prediction modes according to the intra prediction mode corresponding to the weight derivation mode one.
  • the first prediction mode and/or the second prediction mode are intra-frame prediction modes
  • the first prediction mode and/or the second prediction mode are frames corresponding to the weight derivation mode Determined in intra prediction mode.
  • the first prediction mode and/or the second prediction mode may be an intra prediction mode that is on the same straight line or approximately on the same straight line as the weight division line (also referred to as a boundary line).
  • the first prediction mode and/or the second prediction mode may be an intra prediction mode perpendicular to or approximately perpendicular to the weight dividing line.
  • intra prediction modes corresponding to the weight derivation mode there are many types of intra prediction modes corresponding to the weight derivation mode, for example, including intra prediction modes parallel to the boundary line of weights, intra prediction modes perpendicular to the boundary line, and the like.
  • the present application may use a flag to indicate which one of the intra prediction modes corresponding to the weight derivation mode is specifically selected for the first prediction mode and/or the second prediction mode.
  • the first prediction mode is an intra prediction mode
  • use the second flag to indicate the correspondence between the first prediction mode and the intra prediction mode corresponding to the weight derivation mode, for example, the second The flag indicates that the first prediction mode is an intra prediction mode parallel to the boundary line of weights, or indicates that the first prediction mode is an intra prediction mode perpendicular to the boundary line of weights.
  • a third flag is used to indicate the correspondence between the second prediction mode and the intra prediction mode corresponding to the weight derivation mode, for example, the third flag indicates that the second prediction mode is An intra-frame prediction mode parallel to the dividing line of the weight, or indicating that the second prediction mode is an intra-frame prediction mode perpendicular to the dividing line of the weight.
  • the way of determining the first prediction mode and/or the second prediction mode according to the intra prediction mode corresponding to the weight derivation mode in the above method 2 includes but not limited to the following examples:
  • Example 1 if the first prediction mode is an intra prediction mode, obtain the second flag, and determine the intra prediction mode corresponding to the second flag among the intra prediction modes corresponding to the weight derivation mode as the first prediction mode.
  • Example 2 if the second prediction mode is the intra prediction mode, the third flag is obtained, and the intra prediction mode corresponding to the third flag among the intra prediction modes corresponding to the weight derivation mode is determined as the second prediction mode.
  • the intra prediction mode corresponding to the weight derivation mode includes at least one of an intra prediction mode parallel to the boundary line of the weight and an intra prediction mode perpendicular to the boundary line.
  • the intra prediction mode corresponding to the weight derivation mode includes at least one of an intra prediction mode parallel to the boundary line of the weight, an intra prediction mode perpendicular to the boundary line, and a planar mode.
  • the second flag (intra_gpm_idx0) and/or the third flag (intra_gpm_idx1) can be written into the code stream according to the manner shown in Table 8 above.
  • the encoding end writes the second flag and/or the third flag into the code stream in the manner of Table 8 above.
  • the decoding end decodes the code stream, obtains the second flag and/or the third flag, and determines the first prediction mode and/or the second prediction mode according to the second flag and/or the third flag, and then uses the first prediction mode and The second prediction mode and weight determine the prediction value.
  • At least one of K prediction modes is determined according to the weight derivation mode, K preset values are determined according to the K prediction modes, and the K prediction values are weighted to obtain the final prediction value.
  • the encoder derives the mode by determining the weight of the current block; determines K templates according to at least one of the size of the current block and the weight derivation mode; and determines K prediction modes according to the K templates; Determine the prediction value according to the K prediction modes and the weight derivation mode. That is, the present application derives the mode based on the size and/or weight of the current block when determining K templates, so that the determined K templates are more in line with the actual situation, so that when using these K templates to determine the prediction mode, the prediction mode can be improved. The accuracy of the determination, and then use the accurately determined K prediction modes to achieve accurate prediction of the current block and improve the coding effect.
  • sequence numbers of the above-mentioned processes do not mean the order of execution, and the order of execution of the processes should be determined by their functions and internal logic, and should not be used in this application.
  • the implementation of the examples constitutes no limitation.
  • the term "and/or" is only an association relationship describing associated objects, indicating that there may be three relationships. Specifically, A and/or B may mean: A exists alone, A and B exist simultaneously, and B exists alone.
  • the character "/" in this application generally indicates that the contextual objects are an "or" relationship.
  • FIG. 22 is a schematic block diagram of a prediction device provided by an embodiment of the present application, and the prediction device 10 is applied to the above-mentioned video decoder.
  • the forecasting device 10 includes:
  • the decoding unit 11 is used to decode the code stream and determine the weight derivation mode of the current block
  • a template determination unit 12 configured to determine K templates according to at least one of the size of the current block and the weight derivation mode, where K is a positive integer greater than 1;
  • a mode determination unit 13 configured to determine K prediction modes according to the K templates
  • the prediction unit 14 is configured to determine a prediction value according to the K prediction modes and the weight derivation mode.
  • the template determination unit 12 is specifically configured to divide the template of the current block into the K templates according to the weight derivation mode.
  • the template determination unit 12 is specifically configured to divide the template of the current block into M sub-templates, where M is a positive integer greater than or equal to K; according to the weight derivation mode, the M The sub-templates correspond to the K templates.
  • the template determining unit 12 is specifically configured to divide the template of the current block into M sub-templates according to the weight derivation mode.
  • the template determining unit 12 is specifically configured to determine the boundary line of the weight according to the weight derivation mode; and extend the boundary line to the template of the current block, so that the weight of the current block
  • the template is divided into M sub-templates.
  • the template determining unit 12 is specifically configured to extend the boundary line into the template of the current block to obtain an extension line of the boundary line in the template of the current block; using the The extension line divides the template of the current block into M rectangular sub-templates.
  • the template determination unit 12 is specifically configured to divide the upper template of the current block into P sub-templates; and/or, divide the left template of the current block into Q sub-templates; wherein, Both P and Q are integers less than or equal to M, and the sum of P and Q is equal to M.
  • the template determination unit 12 is specifically configured to divide the upper template into P sub-templates along the vertical direction.
  • the template determination unit 12 is specifically configured to divide the upper template into P sub-templates according to a preset number of pixels.
  • the template determination unit 12 is specifically configured to divide the upper template into P sub-templates by taking n columns of pixels as a minimum division unit, where n is a positive integer.
  • the n is determined according to the length of the current block.
  • the template determining unit 12 is specifically configured to divide the left template into Q sub-templates along the horizontal direction.
  • the template determining unit 12 is specifically configured to divide the left template into Q sub-templates according to a preset number of pixels.
  • the template determination unit 12 is specifically configured to divide the left template into Q sub-templates by using m rows of pixels as a minimum division unit, where m is a positive integer.
  • the m is determined according to the width of the current block.
  • the template determining unit 12 is specifically configured to, for the j-th sub-template among the M sub-templates, determine the relationship between the first point of the j-th sub-template and the i-th sub-template according to the weight derivation mode
  • the weight of the prediction mode, the i-th prediction mode is any one of the K prediction modes; according to the weight of the first point in the j-th sub-template about the i-th prediction mode, the The jth sub-template corresponds to the K templates.
  • the template determining unit 12 is specifically configured to determine an angle index and a distance index according to the weight derivation mode; according to the angle index and the distance index, determine that the first point in the jth sub-template is about the first point in the jth sub-template Weights for the i prediction modes.
  • the first point is a point on the boundary line between the jth sub-template and the current block.
  • the first point is a midpoint of the boundary line.
  • the template determining unit 12 is specifically configured to determine a first parameter of the first point according to the angle index, the distance index, and the size of the current block, and the first parameter is used for Determine the weight; determine the weight of the first point with respect to the i-th prediction mode according to the first parameter of the first point.
  • the template determining unit 12 is specifically configured to determine the second parameter of the first point according to the first parameter of the first point; and determine the second parameter of the first point according to the second parameter of the first point.
  • the first point is about the weight of the i-th prediction mode.
  • the template determining unit 12 is specifically configured to determine the i-th prediction mode of the first point according to the first parameter, the first preset value, and the second preset value of the first point. Weights.
  • the weight of the first point with respect to the i-th prediction mode is a first value or a second value.
  • the template determination unit 12 is specifically configured to map the j-th sub-template to the i-th template if the weight of the first point with respect to the i-th prediction mode is greater than a first preset value , the i-th template is one of the K templates.
  • the template determination unit 12 is specifically configured to: if the weight of the first point with respect to the first prediction mode is greater than or equal to a second preset value, then The j-th sub-template is corresponding to the first template; if the weight of the first point with respect to the first prediction mode is less than a second preset value, then the j-th sub-template is corresponding to the second template.
  • the template determining unit 12 is specifically configured to determine the target first correspondence corresponding to the current block from the preset first correspondence corresponding to different block sizes, the first correspondence includes Correspondence between different angle indexes or different weight derivation modes and the K templates; from the target first correspondence, determine the K templates corresponding to the weight derivation modes.
  • the mode determination unit 13 is specifically configured to obtain at least one candidate prediction mode for the i-th prediction mode among the K prediction modes; Predict the i-th template to obtain the predicted sample of the i-th template; determine the cost of the candidate prediction mode according to the predicted value and reconstruction value of the i-th template; determine the cost of the candidate prediction mode according to the at least one candidate prediction mode cost, determine the i-th prediction mode.
  • the prediction unit 14 is specifically configured to determine weights according to the weight derivation mode; determine K prediction values according to the K prediction modes; and weight the K prediction values according to the weights , to get the final predicted value.
  • the mode determining unit 13 is specifically configured to determine K prediction modes according to the K templates if the points included in the K templates are all greater than a preset threshold.
  • the mode determination unit 13 is further configured to determine the K prediction modes according to the weight derivation mode if at least one of the K templates includes points less than a preset threshold.
  • the template determination unit 12 is specifically configured to decode the code stream to obtain a first flag, and the first flag is used to indicate whether to use template matching to derive the prediction mode; if the first flag indicates to use When the prediction mode is derived by template matching, K templates are determined according to at least one of the size of the current block and the weight derivation mode.
  • the mode determination unit 13 is further configured to determine the K prediction modes according to the weight derivation mode if the first flag indicates that the template matching manner is not used to derive the prediction mode.
  • the mode determination unit 13 is specifically configured to determine an angle index according to the weight derivation mode; The corresponding intra prediction mode is determined as at least one of the K prediction modes.
  • the mode determination unit 13 is specifically configured to obtain the intra-frame prediction mode corresponding to the weight derivation mode; according to the The intra prediction mode corresponding to the weight derivation mode is to determine at least one of the K prediction modes.
  • the intra prediction mode corresponding to the weight derivation mode includes at least one of an intra prediction mode parallel to the boundary line of the weight, an intra prediction mode perpendicular to the boundary line, and a planar mode.
  • the prediction unit 14 is specifically configured to determine motion information according to the i-th prediction mode; according to The motion information is used to determine the i-th prediction value; to determine K-1 prediction values according to other prediction modes in the K prediction modes except the i-th prediction mode; to determine the weight according to the weight derivation mode ; Determine the final predicted value according to the ith predicted value, the K-1 predicted values and the weight.
  • the device embodiment and the method embodiment may correspond to each other, and similar descriptions may refer to the method embodiment. To avoid repetition, details are not repeated here.
  • the device 10 shown in FIG. 22 can execute the prediction method at the decoding end of the embodiment of the present application, and the aforementioned and other operations and/or functions of each unit in the device 10 are to realize the above-mentioned prediction method at the decoding end and other methods. For the sake of brevity, the corresponding process will not be repeated here.
  • Fig. 23 is a schematic block diagram of a prediction device provided by an embodiment of the present application, and the prediction device is applied to the above encoder.
  • the prediction device 20 may include:
  • a determining unit 21 configured to determine the weight derivation mode of the current block
  • a template determination unit 22 configured to determine K templates according to at least one of the size of the current block and the weight derivation mode, where K is a positive integer greater than 1;
  • a mode determination unit 23 configured to determine K prediction modes according to the K templates
  • the prediction unit 24 is configured to determine a prediction value according to the K prediction modes and the weight derivation mode.
  • the template determination unit 22 is specifically configured to divide the template of the current block into the K templates according to the weight derivation mode.
  • the template determination unit 22 is specifically configured to divide the template of the current block into M sub-templates, where M is a positive integer greater than or equal to K; according to the weight derivation mode, the M The sub-templates correspond to the K templates.
  • the template determination unit 22 is specifically configured to divide the template of the current block into M sub-templates according to the weight derivation mode.
  • the template determining unit 22 is specifically configured to determine the boundary line of the weight according to the weight derivation mode; and extend the boundary line to the template of the current block, so that the weight of the current block
  • the template is divided into M sub-templates.
  • the template determination unit 22 is specifically configured to extend the boundary line into the template of the current block to obtain an extension line of the boundary line in the template of the current block; using the The extension line divides the template of the current block into M rectangular sub-templates.
  • the template determination unit 22 is specifically configured to divide the upper template of the current block into P sub-templates; and/or, divide the left template of the current block into Q sub-templates; wherein, Both P and Q are integers less than or equal to M, and the sum of P and Q is equal to M.
  • the template determination unit 22 is specifically configured to divide the upper template into P sub-templates along the vertical direction.
  • the template determining unit 22 is specifically configured to divide the upper template into P sub-templates according to a preset number of pixels.
  • the template determining unit 22 is specifically configured to divide the upper template into P sub-templates by taking n columns of pixels as a minimum division unit, where n is a positive integer.
  • the n is determined according to the length of the current block.
  • the template determining unit 22 is specifically configured to divide the left template into Q sub-templates along the horizontal direction.
  • the template determining unit 22 is specifically configured to divide the left template into Q sub-templates according to a preset number of pixels.
  • the template determination unit 22 is specifically configured to divide the left template into Q sub-templates by using m rows of pixels as a minimum division unit, and m is a positive integer.
  • the m is determined according to the width of the current block.
  • the template determining unit 22 is specifically configured to, for the j-th sub-template among the M sub-templates, determine the relationship between the first point of the j-th sub-template and the i-th sub-template according to the weight derivation mode
  • the weight of the prediction mode, the i-th prediction mode is any one of the K prediction modes; according to the weight of the first point in the j-th sub-template about the i-th prediction mode, the The jth sub-template corresponds to the K templates.
  • the template determining unit 22 is specifically configured to determine an angle index and a distance index according to the weight derivation mode; according to the angle index and the distance index, determine that the first point in the jth sub-template is about the first point in the jth sub-template Weights for the i prediction modes.
  • the first point is a point on the boundary line between the jth sub-template and the current block.
  • the first point is a midpoint of the boundary line.
  • the template determining unit 22 is specifically configured to determine a first parameter of the first point according to the angle index, the distance index, and the size of the current block, and the first parameter is used for Determine the weight; determine the weight of the first point with respect to the i-th prediction mode according to the first parameter of the first point.
  • the template determining unit 22 is specifically configured to determine the second parameter of the first point according to the first parameter of the first point; and determine the second parameter of the first point according to the second parameter of the first point.
  • the first point is about the weight of the i-th prediction mode.
  • the template determination unit 22 is specifically configured to determine the i-th prediction mode of the first point according to the first parameter, the first preset value, and the second preset value of the first point. Weights.
  • the weight of the first point with respect to the i-th prediction mode is the first value or the second value.
  • the template determining unit 22 is specifically configured to map the j-th sub-template to the i-th template if the weight of the first point with respect to the i-th prediction mode is greater than a first preset value , the i-th template is one of the K templates.
  • the template determination unit 22 is specifically configured to: if the weight of the first point with respect to the first prediction mode is greater than or equal to the second preset value, then The j-th sub-template is corresponding to the first template; if the weight of the first point with respect to the first prediction mode is less than a second preset value, then the j-th sub-template is corresponding to the second template.
  • the template determining unit 22 is specifically configured to determine the target first correspondence corresponding to the current block from the preset first correspondence corresponding to different block sizes, the first correspondence includes Correspondence between different angle indexes or different weight derivation modes and the K templates; from the target first correspondence, determine the K templates corresponding to the weight derivation modes.
  • the mode determination unit 23 is specifically configured to acquire at least one candidate prediction mode for the i-th prediction mode among the K prediction modes; Predict the i-th template to obtain the predicted sample of the i-th template; determine the cost of the candidate prediction mode according to the predicted value and reconstruction value of the i-th template; determine the cost of the candidate prediction mode according to the at least one candidate prediction mode cost, determine the i-th prediction mode.
  • the prediction unit 24 is specifically configured to determine weights according to the weight derivation mode; determine K prediction values according to the K prediction modes; and weight the K prediction values according to the weights , to get the final predicted value.
  • the mode determining unit 23 is specifically configured to determine K prediction modes according to the K templates if the points included in the K templates are all greater than a preset threshold.
  • the mode determination unit 23 is further configured to determine the K prediction modes according to the weight derivation mode if at least one of the K templates includes points less than a preset threshold.
  • the mode determining unit 23 is further configured to write a first flag into the code stream, where the first flag is used to indicate whether to use template matching to derive the prediction mode.
  • the mode determination unit 23 is specifically configured to determine an angle index according to the weight derivation mode; An intra prediction mode is determined as at least one of the K prediction modes.
  • the mode determination unit 23 is specifically configured to acquire the intra-frame prediction mode corresponding to the weight derivation mode; according to the The intra prediction mode corresponding to the weight derivation mode is to determine at least one of the K prediction modes.
  • the intra prediction mode corresponding to the weight derivation mode includes at least one of an intra prediction mode parallel to the boundary line of the weight, an intra prediction mode perpendicular to the boundary line, and a planar mode.
  • the prediction unit 24 is specifically configured to determine motion information according to the i-th prediction mode; according to The motion information is used to determine the i-th prediction value; to determine K-1 prediction values according to other prediction modes in the K prediction modes except the i-th prediction mode; to determine the weight according to the weight derivation mode ; Determine the final predicted value according to the ith predicted value, the K-1 predicted values and the weight.
  • the device embodiment and the method embodiment may correspond to each other, and similar descriptions may refer to the method embodiment. To avoid repetition, details are not repeated here.
  • the device 20 shown in FIG. 23 may correspond to the corresponding subject in the prediction method at the encoding end of the embodiment of the present application, and the aforementioned and other operations and/or functions of each unit in the device 20 are respectively to realize the prediction at the encoding end
  • the corresponding processes in each method, such as the method will not be repeated here.
  • the functional unit may be implemented in the form of hardware, may also be implemented by instructions in the form of software, and may also be implemented by a combination of hardware and software units.
  • each step of the method embodiment in the embodiment of the present application can be completed by an integrated logic circuit of the hardware in the processor and/or instructions in the form of software, and the steps of the method disclosed in the embodiment of the present application can be directly embodied as hardware
  • the decoding processor is executed, or the combination of hardware and software units in the decoding processor is used to complete the execution.
  • the software unit may be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, and registers.
  • the storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps in the above method embodiments in combination with its hardware.
  • Fig. 24 is a schematic block diagram of an electronic device provided by an embodiment of the present application.
  • the electronic device 30 may be the video encoder or video decoder described in the embodiment of the present application, and the electronic device 30 may include:
  • a memory 33 and a processor 32 the memory 33 is used to store a computer program 34 and transmit the program code 34 to the processor 32 .
  • the processor 32 can call and run the computer program 34 from the memory 33 to implement the method in the embodiment of the present application.
  • the processor 32 can be used to execute the steps in the above-mentioned method 200 according to the instructions in the computer program 34 .
  • the processor 32 may include, but is not limited to:
  • DSP Digital Signal Processor
  • ASIC Application Specific Integrated Circuit
  • FPGA Field Programmable Gate Array
  • the memory 33 includes but is not limited to:
  • non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electronically programmable Erase Programmable Read-Only Memory (Electrically EPROM, EEPROM) or Flash.
  • the volatile memory can be Random Access Memory (RAM), which acts as external cache memory.
  • RAM Static Random Access Memory
  • SRAM Static Random Access Memory
  • DRAM Dynamic Random Access Memory
  • Synchronous Dynamic Random Access Memory Synchronous Dynamic Random Access Memory
  • SDRAM double data rate synchronous dynamic random access memory
  • Double Data Rate SDRAM, DDR SDRAM double data rate synchronous dynamic random access memory
  • Enhanced SDRAM, ESDRAM enhanced synchronous dynamic random access memory
  • SLDRAM synchronous connection dynamic random access memory
  • Direct Rambus RAM Direct Rambus RAM
  • the computer program 34 can be divided into one or more units, and the one or more units are stored in the memory 33 and executed by the processor 32 to complete the present application.
  • the one or more units may be a series of computer program instruction segments capable of accomplishing specific functions, and the instruction segments are used to describe the execution process of the computer program 34 in the electronic device 30 .
  • the electronic device 30 may also include:
  • a transceiver 33 the transceiver 33 can be connected to the processor 32 or the memory 33 .
  • the processor 32 can control the transceiver 33 to communicate with other devices, specifically, can send information or data to other devices, or receive information or data sent by other devices.
  • Transceiver 33 may include a transmitter and a receiver.
  • the transceiver 33 may further include antennas, and the number of antennas may be one or more.
  • bus system includes not only a data bus, but also a power bus, a control bus and a status signal bus.
  • Fig. 25 is a schematic block diagram of a video encoding and decoding system provided by an embodiment of the present application.
  • the video codec system 40 may include: a video encoder 41 and a video decoder 42, wherein the video encoder 41 is used to execute the video encoding method involved in the embodiment of the present application, and the video decoder 42 is used to execute The video decoding method involved in the embodiment of the present application.
  • the present application also provides a computer storage medium, on which a computer program is stored, and when the computer program is executed by a computer, the computer can execute the methods of the above method embodiments.
  • the embodiments of the present application further provide a computer program product including instructions, and when the instructions are executed by a computer, the computer executes the methods of the foregoing method embodiments.
  • the present application also provides a code stream, which is generated according to the above encoding method.
  • the computer program product includes one or more computer instructions.
  • the computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable device.
  • the computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, e.g. (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) to another website site, computer, server or data center.
  • the computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or a data center integrated with one or more available media.
  • the available medium may be a magnetic medium (such as a floppy disk, a hard disk, or a magnetic tape), an optical medium (such as a digital video disc (digital video disc, DVD)), or a semiconductor medium (such as a solid state disk (solid state disk, SSD)), etc.
  • the disclosed systems, devices and methods may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components can be combined or can be Integrate into another system, or some features may be ignored, or not implemented.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.
  • a unit described as a separate component may or may not be physically separated, and a component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Compression Of Band Width Or Redundancy In Fax (AREA)

Abstract

一种预测方法、装置、设备、系统、及存储介质，该方法包括确定当前块的权重导出模式；根据当前块的大小和权重导出模式中的至少一个，确定K个模板；根据K个模板，确定K个预测模式；根据K个预测模式和权重导出模式，确定预测值。在确定K个模板时是基于当前块的大小和/或权重导出模式的，使得确定出的K个模板更加符合实际情况，这样使用这K个模板确定预测模式时，可以提高预测模式的确定准确性，进而使用准确确定的K个预测模式实现当前块的准确预测，提升编码效果。

Description

预测方法、装置、设备、系统、及存储介质 技术领域
本申请涉及视频编解码技术领域，尤其涉及一种预测方法、装置、设备、系统、及存储介质。
背景技术
数字视频技术可以并入多种视频装置中，例如数字电视、智能手机、计算机、电子阅读器或视频播放器等。随着视频技术的发展，视频数据所包括的数据量较大，为了便于视频数据的传输，视频装置执行视频压缩技术，以使视频数据更加有效的传输或存储。
由于视频中存在时间或空间冗余，通过预测可以消除或降低视频中的冗余，提高压缩效率。在预测时，首先确定预测模式，例如通过模板匹配方式确定当前块的第一预测模式和第二预测模式。但是，目前对模板的划分不够详细，使得根据模板确定第一预测模式和第二预测模式时，确定的预测模式不准确，进而导出压缩效果差。
发明内容
本申请实施例提供了一种预测方法、装置、设备、系统、及存储介质，提高了模板的划分准确性，进而提升压缩性能。
第一方面，本申请提供了一种预测方法，应用于解码器，包括：
解码码流，确定当前块的权重导出模式；
根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
根据所述K个模板，确定K个预测模式；
根据所述K个预测模式和所述权重导出模式，确定预测值。
第二方面，本申请实施例提供一种预测方法，包括：
确定当前块的权重导出模式；
根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
根据所述K个模板，确定K个预测模式；
根据所述K个预测模式和所述权重导出模式，确定预测值。
第三方面，本申请提供了一种预测装置，用于执行上述第一方面或其各实现方式中的方法。具体地，该预测装置包括用于执行上述第一方面或其各实现方式中的方法的功能单元。
第四方面，本申请提供了一种预测装置，用于执行上述第二方面或其各实现方式中的方法。具体地，该预测装置包括用于执行上述第二方面或其各实现方式中的方法的功能单元。
第五方面，提供了一种视频解码器，包括处理器和存储器。该存储器用于存储计算机程序，该处理器用于调用并运行该存储器中存储的计算机程序，以执行上述第一方面或其各实现方式中的方法。
第六方面，提供了一种视频编码器，包括处理器和存储器。该存储器用于存储计算机程序，该处理器用于调用并运行该存储器中存储的计算机程序，以执行上述第二方面或其各实现方式中的方法。
第七方面，提供了一种视频编解码系统，包括视频编码器和视频解码器。视频解码器用于执行上述第一方面或其各实现方式中的方法，视频编码器用于执行上述第二方面或其各实现方式中的方法。
第八方面，提供了一种芯片，用于实现上述第一方面至第二方面中的任一方面或其各实现方式中的方法。具体地，该芯片包括：处理器，用于从存储器中调用并运行计算机程序，使得安装有该芯片的设备执行如上述第一方面至第二方面中的任一方面或其各实现方式中的方法。
第九方面，提供了一种计算机可读存储介质，用于存储计算机程序，该计算机程序使得计算机执行上述第一方面至第二方面中的任一方面或其各实现方式中的方法。
第十方面，提供了一种计算机程序产品，包括计算机程序指令，该计算机程序指令使得计算机执行上述第一方面至第二方面中的任一方面或其各实现方式中的方法。
第十一方面，提供了一种计算机程序，当其在计算机上运行时，使得计算机执行上述第一方面至第二方面中的任一方面或其各实现方式中的方法。
第十二方面，提供了一种码流，码流是基于上述第二方面的方法生成的。
基于以上技术方案，通过确定当前块的权重导出模式；根据当前块的大小和权重导出模式中的至少一个，确定K个模板；根据K个模板，确定K个预测模式；根据K个预测模式和权重导出模式，确定预测值。即本申请在确定K个模板时是基于当前块的大小和/或权重导出模式的，使得确定出的K个模板更加符合实际情况，这样使用这K个模板确定预测模式时，可以提高预测模式的确定准确性，进而使用准确确定的K个预测模式实现当前块的准确预测，提升编码效果。
附图说明
图1为本申请实施例涉及的一种视频编解码系统的示意性框图；
图2是本申请实施例涉及的视频编码器的示意性框图；
图3是本申请实施例涉及的视频解码器的示意性框图；
图4为权重分配示意图；
图5为权重分配示意图；
图6A为帧间预测的示意图；
图6B为加权帧间预测的示意图；
图7A为帧内预测的示意图；
图7B为帧内预测的示意图；
图8A-8I为帧内预测的示意图；
图9为帧内预测模式的示意图；
图10为帧内预测模式的示意图；
图11为帧内预测模式的示意图；
图12为加权帧内预测的示意图；
图13为模板匹配示意图；
图14为本申请一实施例提供的预测方法流程示意图；
图15为使用两种预测模式对当前块进行预测时的示意图；
图16为一种模板划分示意图；
图17A-17G为一种模板划分示意图；
图18A-18D为另一种模板划分示意图；
图19为一种模板尺寸示意图；
图20A为一种权重分配示意图；
图20B为另一种权重分配示意图；
图21为本申请实一施例提供的预测方法流程示意图；
图22是本申请一实施例提供的预测装置的示意性框图；
图23是本申请一实施例提供的预测装置的示意性框图；
图24是本申请实施例提供的电子设备的示意性框图；
图25是本申请实施例提供的视频编解码系统的示意性框图。
具体实施方式
本申请可应用于图像编解码领域、视频编解码领域、硬件视频编解码领域、专用电路视频编解码领域、实时视频编解码领域等。例如，本申请的方案可结合至音视频编码标准(audio video coding standard，简称AVS)，例如，H.264/音视频编码(audio video coding，简称AVC)标准，H.265/高效视频编码(high efficiency video coding，简称HEVC)标准以及H.266/多功能视频编码(versatile video coding，简称VVC)标准。或者，本申请的方案可结合至其它专属或行业标准而操作，所述标准包含ITU-TH.261、ISO/IECMPEG-1Visual、ITU-TH.262或ISO/IECMPEG-2Visual、ITU-TH.263、ISO/IECMPEG-4Visual，ITU-TH.264(还称为ISO/IECMPEG-4AVC)，包含可分级视频编解码(SVC)及多视图视频编解码(MVC)扩展。应理解，本申请的技术不限于任何特定编解码标准或技术。
为了便于理解，首先结合图1对本申请实施例涉及的视频编解码系统进行介绍。
图1为本申请实施例涉及的一种视频编解码系统的示意性框图。需要说明的是，图1只是一种示例，本申请实施例的视频编解码系统包括但不限于图1所示。如图1所示，该视频编解码系统100包含编码设备110和解码设备120。其中编码设备用于对视频数据进行编码(可以理解成压缩)产生码流，并将码流传输给解码设备。解码设备对编码设备编码产生的码流进行解码，得到解码后的视频数据。
本申请实施例的编码设备110可以理解为具有视频编码功能的设备，解码设备120可以理解为具有视频解码功能的设备，即本申请实施例对编码设备110和解码设备120包括更广泛的装置，例如包含智能手机、台式计算机、移动计算装置、笔记本(例如，膝上型)计算机、平板计算机、机顶盒、电视、相机、显示装置、数字媒体播放器、视频游戏控制台、车载计算机等。
在一些实施例中，编码设备110可以经由信道130将编码后的视频数据(如码流)传输给解码设备120。信道130可以包括能够将编码后的视频数据从编码设备110传输到解码设备120的一个或多个媒体和/或装置。
在一个实例中，信道130包括使编码设备110能够实时地将编码后的视频数据直接发射到解码设备120的一个或多个通信媒体。在此实例中，编码设备110可根据通信标准来调制编码后的视频数据，且将调制后的视频数据发射到解码设备120。其中通信媒体包含无线通信媒体，例如射频频谱，可选的，通信媒体还可以包含有线通信媒体，例如一根或多根物理传输线。
在另一实例中，信道130包括存储介质，该存储介质可以存储编码设备110编码后的视频数据。存储介质包含多种本地存取式数据存储介质，例如光盘、DVD、快闪存储器等。在该实例中，解码设备120可从该存储介质中获取编码后的视频数据。
在另一实例中，信道130可包含存储服务器，该存储服务器可以存储编码设备110编码后的视频数据。在此实例中，解码设备120可以从该存储服务器中下载存储的编码后的视频数据。可选的，该存储服务器可以存储编码后的视频数据且可以将该编码后的视频数据发射到解码设备120，例如web服务器(例如，用于网站)、文件传送协议(FTP)服务器等。
一些实施例中，编码设备110包含视频编码器112及输出接口113。其中，输出接口113可以包含调制器/解调器(调制解调器)和/或发射器。
在一些实施例中，编码设备110除了包括视频编码器112和输入接口113外，还可以包括视频源111。
视频源111可包含视频采集装置(例如，视频相机)、视频存档、视频输入接口、计算机图形系统中的至少一个，其中，视频输入接口用于从视频内容提供者处接收视频数据，计算机图形系统用于产生视频数据。
视频编码器112对来自视频源111的视频数据进行编码，产生码流。视频数据可包括一个或多个图像(picture)或图像序列(sequence of pictures)。码流以比特流的形式包含了图像或图像序列的编码信息。编码信息可以包含编码 图像数据及相关联数据。相关联数据可包含序列参数集(sequence parameter set，简称SPS)、图像参数集(picture parameter set，简称PPS)及其它语法结构。SPS可含有应用于一个或多个序列的参数。PPS可含有应用于一个或多个图像的参数。语法结构是指码流中以指定次序排列的零个或多个语法元素的集合。
视频编码器112经由输出接口113将编码后的视频数据直接传输到解码设备120。编码后的视频数据还可存储于存储介质或存储服务器上，以供解码设备120后续读取。
在一些实施例中，解码设备120包含输入接口121和视频解码器122。
在一些实施例中，解码设备120除包括输入接口121和视频解码器122外，还可以包括显示装置123。
其中，输入接口121包含接收器及/或调制解调器。输入接口121可通过信道130接收编码后的视频数据。
视频解码器122用于对编码后的视频数据进行解码，得到解码后的视频数据，并将解码后的视频数据传输至显示装置123。
显示装置123显示解码后的视频数据。显示装置123可与解码设备120整合或在解码设备120外部。显示装置123可包括多种显示装置，例如液晶显示器(LCD)、等离子体显示器、有机发光二极管(OLED)显示器或其它类型的显示装置。
此外，图1仅为实例，本申请实施例的技术方案不限于图1，例如本申请的技术还可以应用于单侧的视频编码或单侧的视频解码。
下面对本申请实施例涉及的视频编码框架进行介绍。
图2是本申请实施例涉及的视频编码器的示意性框图。应理解，该视频编码器200可用于对图像进行有损压缩(lossy compression)，也可用于对图像进行无损压缩(lossless compression)。该无损压缩可以是视觉无损压缩(visually lossless compression)，也可以是数学无损压缩(mathematically lossless compression)。
该视频编码器200可应用于亮度色度(YCbCr，YUV)格式的图像数据上。例如，YUV比例可以为4:2:0、4:2:2或者4:4:4，Y表示明亮度(Luma)，Cb(U)表示蓝色色度，Cr(V)表示红色色度，U和V表示为色度(Chroma)用于描述色彩及饱和度。例如，在颜色格式上，4:2:0表示每4个像素有4个亮度分量，2个色度分量(YYYYCbCr)，4:2:2表示每4个像素有4个亮度分量，4个色度分量(YYYYCbCrCbCr)，4:4:4表示全像素显示(YYYYCbCrCbCrCbCrCbCr)。
例如，该视频编码器200读取视频数据，针对视频数据中的每帧图像，将一帧图像划分成若干个编码树单元(coding tree unit，CTU)，在一些例子中，CTB可被称作“树型块”、“最大编码单元”(Largest Coding unit，简称LCU)或“编码树型块”(coding tree block，简称CTB)。每一个CTU可以与图像内的具有相等大小的像素块相关联。每一像素可对应一个亮度(luminance或luma)采样及两个色度(chrominance或chroma)采样。因此，每一个CTU可与一个亮度采样块及两个色度采样块相关联。一个CTU大小例如为128×128、64×64、32×32等。一个CTU又可以继续被划分成若干个编码单元(Coding Unit，CU)进行编码，CU可以为矩形块也可以为方形块。CU可以进一步划分为预测单元(prediction Unit，简称PU)和变换单元(transform unit，简称TU)，进而使得编码、预测、变换分离，处理的时候更灵活。在一种示例中，CTU以四叉树方式划分为CU，CU以四叉树方式划分为TU、PU。
视频编码器及视频解码器可支持各种PU大小。假定特定CU的大小为2N×2N，视频编码器及视频解码器可支持2N×2N或N×N的PU大小以用于帧内预测，且支持2N×2N、2N×N、N×2N、N×N或类似大小的对称PU以用于帧间预测。视频编码器及视频解码器还可支持2N×nU、2N×nD、nL×2N及nR×2N的不对称PU以用于帧间预测。
在一些实施例中，如图2所示，该视频编码器200可包括：预测单元210、残差单元220、变换/量化单元230、反变换/量化单元240、重建单元250、环路滤波单元260、解码图像缓存270和熵编码单元280。需要说明的是，视频编码器200可包含更多、更少或不同的功能组件。
可选的，在本申请中，当前块(current block)可以称为当前编码单元(CU)或当前预测单元(PU)等。预测块也可称为预测图像块或图像预测块，重建图像块也可称为重建块或图像重建图像块。
在一些实施例中，预测单元210包括帧间预测单元211和帧内估计单元212。由于视频的一个帧中的相邻像素之间存在很强的相关性，在视频编解码技术中使用帧内预测的方法消除相邻像素之间的空间冗余。由于视频中的相邻帧之间存在着很强的相似性，在视频编解码技术中使用帧间预测方法消除相邻帧之间的时间冗余，从而提高编码效率。
帧间预测单元211可用于帧间预测，帧间预测可以包括运动估计(motion estimation)和运动补偿(motion compensation)，可以参考不同帧的图像信息，帧间预测使用运动信息从参考帧中找到参考块，根据参考块生成预测块，用于消除时间冗余；帧间预测所使用的帧可以为P帧和/或B帧，P帧指的是向前预测帧，B帧指的是双向预测帧。帧间预测使用运动信息从参考帧中找到参考块，根据参考块生成预测块。运动信息包括参考帧所在的参考帧列表，参考帧索引，以及运动矢量。运动矢量可以是整像素的或者是分像素的，如果运动矢量是分像素的，那么需要在参考帧中使用插值滤波做出所需的分像素的块，这里把根据运动矢量找到的参考帧中的整像素或者分像素的块叫参考块。有的技术会直接把参考块作为预测块，有的技术会在参考块的基础上再处理生成预测块。在参考块的基础上再处理生成预测块也可以理解为把参考块作为预测块然后再在预测块的基础上处理生成新的预测块。
帧内估计单元212只参考同一帧图像的信息，预测当前码图像块内的像素信息，用于消除空间冗余。帧内预测所使用的帧可以为I帧。
帧内预测有多种预测模式，以国际数字视频编码标准H系列为例，H.264/AVC标准有8种角度预测模式和1种非角度预测模式，H.265/HEVC扩展到33种角度预测模式和2种非角度预测模式。HEVC使用的帧内预测模式有平面模式(Planar)、DC和33种角度模式，共35种预测模式。VVC使用的帧内模式有Planar、DC和65种角度模式，共67种预测模式。
需要说明的是，随着角度模式的增加，帧内预测将会更加精确，也更加符合对高清以及超高清数字视频发展的需求。
残差单元220可基于CU的像素块及CU的PU的预测块来产生CU的残差块。举例来说，残差单元220可产生 CU的残差块，使得残差块中的每一采样具有等于以下两者之间的差的值：CU的像素块中的采样，及CU的PU的预测块中的对应采样。
变换/量化单元230可量化变换系数。变换/量化单元230可基于与CU相关联的量化参数(QP)值来量化与CU的TU相关联的变换系数。视频编码器200可通过调整与CU相关联的QP值来调整应用于与CU相关联的变换系数的量化程度。
反变换/量化单元240可分别将逆量化及逆变换应用于量化后的变换系数，以从量化后的变换系数重建残差块。
重建单元250可将重建后的残差块的采样加到预测单元210产生的一个或多个预测块的对应采样，以产生与TU相关联的重建图像块。通过此方式重建CU的每一个TU的采样块，视频编码器200可重建CU的像素块。
环路滤波单元260用于对反变换与反量化后的像素进行处理，弥补失真信息，为后续编码像素提供更好的参考，例如可执行消块滤波操作以减少与CU相关联的像素块的块效应。
在一些实施例中，环路滤波单元260包括去块滤波单元和样点自适应补偿/自适应环路滤波(SAO/ALF)单元，其中去块滤波单元用于去方块效应，SAO/ALF单元用于去除振铃效应。
解码图像缓存270可存储重建后的像素块。帧间预测单元211可使用含有重建后的像素块的参考图像来对其它图像的PU执行帧间预测。另外，帧内估计单元212可使用解码图像缓存270中的重建后的像素块来对在与CU相同的图像中的其它PU执行帧内预测。
熵编码单元280可接收来自变换/量化单元230的量化后的变换系数。熵编码单元280可对量化后的变换系数执行一个或多个熵编码操作以产生熵编码后的数据。
图3是本申请实施例涉及的视频解码器的示意性框图。
如图3所示，视频解码器300包含：熵解码单元310、预测单元320、反量化/变换单元330、重建单元340、环路滤波单元350及解码图像缓存360。需要说明的是，视频解码器300可包含更多、更少或不同的功能组件。
视频解码器300可接收码流。熵解码单元310可解析码流以从码流提取语法元素。作为解析码流的一部分，熵解码单元310可解析码流中的经熵编码后的语法元素。预测单元320、反量化/变换单元330、重建单元340及环路滤波单元350可根据从码流中提取的语法元素来解码视频数据，即产生解码后的视频数据。
在一些实施例中，预测单元320包括帧内估计单元322和帧间预测单元321。
帧内估计单元322可执行帧内预测以产生PU的预测块。帧内估计单元322可使用帧内预测模式以基于空间相邻PU的像素块来产生PU的预测块。帧内估计单元322还可根据从码流解析的一个或多个语法元素来确定PU的帧内预测模式。
帧间预测单元321可根据从码流解析的语法元素来构造第一参考图像列表(列表0)及第二参考图像列表(列表1)。此外，如果PU使用帧间预测编码，则熵解码单元310可解析PU的运动信息。帧间预测单元321可根据PU的运动信息来确定PU的一个或多个参考块。帧间预测单元321可根据PU的一个或多个参考块来产生PU的预测块。
反量化/变换单元330可逆量化(即，解量化)与TU相关联的变换系数。反量化/变换单元330可使用与TU的CU相关联的QP值来确定量化程度。
在逆量化变换系数之后，反量化/变换单元330可将一个或多个逆变换应用于逆量化变换系数，以便产生与TU相关联的残差块。
重建单元340使用与CU的TU相关联的残差块及CU的PU的预测块以重建CU的像素块。例如，重建单元340可将残差块的采样加到预测块的对应采样以重建CU的像素块，得到重建图像块。
环路滤波单元350可执行消块滤波操作以减少与CU相关联的像素块的块效应。
视频解码器300可将CU的重建图像存储于解码图像缓存360中。视频解码器300可将解码图像缓存360中的重建图像作为参考图像用于后续预测，或者，将重建图像传输给显示装置呈现。
视频编解码的基本流程如下：在编码端，将一帧图像划分成块，针对当前块，预测单元210使用帧内预测或帧间预测产生当前块的预测块。残差单元220可基于预测块与当前块的原始块计算残差块，即预测块和当前块的原始块的差值，该残差块也可称为残差信息。该残差块经由变换/量化单元230变换与量化等过程，可以去除人眼不敏感的信息，以消除视觉冗余。可选的，经过变换/量化单元230变换与量化之前的残差块可称为时域残差块，经过变换/量化单元230变换与量化之后的时域残差块可称为频率残差块或频域残差块。熵编码单元280接收到变化量化单元230输出的量化后的变化系数，可对该量化后的变化系数进行熵编码，输出码流。例如，熵编码单元280可根据目标上下文模型以及二进制码流的概率信息消除字符冗余。
在解码端，熵解码单元310可解析码流得到当前块的预测信息、量化系数矩阵等，预测单元320基于预测信息对当前块使用帧内预测或帧间预测产生当前块的预测块。反量化/变换单元330使用从码流得到的量化系数矩阵，对量化系数矩阵进行反量化、反变换得到残差块。重建单元340将预测块和残差块相加得到重建块。重建块组成重建图像，环路滤波单元350基于图像或基于块对重建图像进行环路滤波，得到解码图像。编码端同样需要和解码端类似的操作获得解码图像。该解码图像也可以称为重建图像，重建图像可以为后续的帧作为帧间预测的参考帧。
需要说明的是，编码端确定的块划分信息，以及预测、变换、量化、熵编码、环路滤波等模式信息或者参数信息等在必要时携带在码流中。解码端通过解析码流及根据已有信息进行分析确定与编码端相同的块划分信息，预测、变换、量化、熵编码、环路滤波等模式信息或者参数信息，从而保证编码端获得的解码图像和解码端获得的解码图像相同。
上述是基于块的混合编码框架下的视频编解码器的基本流程，随着技术的发展，该框架或流程的一些模块或步骤可能会被优化，本申请适用于该基于块的混合编码框架下的视频编解码器的基本流程，但不限于该框架及流程。
在一些实施例中，当前块(current block)可以是当前编码单元(CU)或当前预测单元(PU)等。由于并行处理 的需要，图像可以被划分成片slice等，同一个图像中的片slice可以并行处理，也就是说它们之间没有数据依赖。而“帧”是一种常用的说法，一般可以理解为一帧是一个图像。在申请中所述帧也可以替换为图像或slice等。
目前正在制定中的多功能视频编码(Versatile Video Coding，VVC)视频编解码标准中，有一个叫做几何划分预测模式(GeometricpartitioningMode，GPM)的帧间预测模式。目前正在制定中的视频编解码标准(Audio Video coding Standard，AVS)视频编解码标准中，有一个叫做角度加权预测模式(Angular Weightedprediction，AWP)的帧间预测模式。这两种模式虽然名称不同、具体的实现形式不同、但原理上有共通之处。
需要说明的是，传统的单向预测只是查找一个与当前块大小相同的参考块，而传统的双向预测使用两个与当前块大小相同的参考块，且预测块内每个点的像素值为两个参考块对应位置的平均值，即每一个参考块的所有点都占50％的比例。双向加权预测使得两个参考块的比例可以不同，如第一个参考块中所有点都占75％的比例，第二个参考块中所有点都占25％的比例。但同一个参考块中的所有点的比例都相同。其他一些优化方式比如采用解码端运动矢量修正(Decoder sideMotion Vector Refinement，DMVR)技术、双向光流(Bi-directional Optical Flow，BIO)等会使参考像素或预测像素产生一些变化，而且GPM或AWP也使用两个与当前块大小相同的参考块，但某些像素位置100％使用第一个参考块对应位置的像素值，某些像素位置100％使用第二个参考块对应位置的像素值，而在交界区域，按一定比例使用这两个参考块对应位置的像素值。具体这些权重如何分配，由GPM或AWP的预测模式决定，或者也可以认为GPM或AWP使用两个与当前块大小不相同的参考块，即各取所需的一部分作为参考块。即将权重不为0的部分作为参考块，而将权重为0的部分剔除出来。
示例性地，图4为权重分配示意图，如图4所示，其示出了本申请实施例提供的一种GPM在64×64的当前块上的多种划分模式的权重分配示意图，其中，GPM存在有64种划分模式。图5为权重分配示意图，如图5所示，其示出了本申请实施例提供的一种AWP在64×64的当前块上的多种划分模式的权重分配示意图，其中，AWP存在有56种划分模式。无论是图4还是图5，每一种划分模式下，黑色区域表示第一个参考块对应位置的权重值为0％，白色区域表示第一个参考块对应位置的权重值为100％，灰色区域则按颜色深浅的不同表示第一个参考块对应位置的权重值为大于0％且小于100％的某一个权重值，第二个参考块对应位置的权重值则为100％减去第一个参考块对应位置的权重值。
GPM和AWP的权重导出方法不同。GPM根据每种模式确定角度及偏移量，而后计算出每个模式的权重矩阵。AWP首先做出一维的权重的线，然后使用类似于帧内角度预测的方法将一维的权重的线铺满整个矩阵。
应理解，早期的编解码技术中只存在矩形的划分方式，无论是CU、PU还是变换单元(Transform Unit，TU)的划分。而GPM或AWP均在没有划分的情况下实现了预测的非矩形的划分效果。GPM和AWP使用了两个参考块的权重的蒙版(mask)，即上述的权重图。这个蒙版确定了两个参考块在产生预测块时的权重，或者可以简单地理解为预测块的一部分位置来自于第一个参考块一部分位置来自于第二个参考块，而过渡区域(blending area)用两个参考块的对应位置加权得到，从而使过渡更平滑。GPM和AWP没有按划分线把当前块划分成两个CU或PU，于是在预测之后的残差的变换、量化、反变换、反量化等也都是将当前块作为一个整体来处理。
GPM使用权重矩阵模拟了几何形状的划分，更确切地说是模拟了预测的划分。而要实施GPM，除了权重矩阵还需要2个预测值，每个预测值由1个单向运动信息确定。这2个单向运动信息来自于一个运动信息候选列表，例如来自merge运动信息候选列表(mergeCandList)。GPM在码流中使用两个索引从mergeCandList中确定2个单向运动信息。
帧间预测使用运动信息(motion information)来表示“运动”。基本的运动信息包含参考帧(reference frame)(或者叫参考图像(reference picture))的信息和运动矢量(MV,motion vector)的信息。常用的双向预测，使用2个参考块对当前块进行预测。2个参考块可以使用一个前向的参考块和一个后向的参考块。可选的，也允许2个都是前向或2个都是后向。所谓前向指参考帧对应的时刻在当前帧之前，后向指参考帧对应的时刻在当前帧之后。或者说前向指参考帧在视频中的位置位于当前帧之前，后向指参考帧在视频中的位置位于当前帧之后。或者说前向指参考帧的POC(picture order count)小于当前帧的POC，后向指参考帧的POC大于当前帧的POC。为了能使用双向预测，自然需要能找到2个参考块，那么就需要2组参考帧的信息和运动矢量的信息。可以把它们每一组理解为一个单向运动信息，而把这2组组合到一起就形成了一个双向运动信息。在具体实现时，单向运动信息和双向运动信息可以使用相同的数据结构，只是双向运动信息的2组参考帧的信息和运动矢量的信息都有效，而单向运动信息的其中一组参考帧的信息和运动矢量的信息是无效的。
在一些实施例中，支持2个参考帧列表，记为RPL0，RPL1，其中RPL是Reference Picture List的简写。在一些实施例中，P slice只可以使用RPL0，B slice可以使用RPL0和RPL1。对一个slice，每个参考帧列表中有若干参考帧，编解码器通过参考帧索引来找到某一个参考帧。在一些实施例中，用参考帧索引和运动矢量来表示运动信息。如对上述的双向运动信息，使用参考帧列表0对应的参考帧索引refIdxL0，以及参考帧列表0对应的运动矢量mvL0，参考帧列表1对应的参考帧索引refIdxL1，以及参考帧列表1对应的运动矢量mvL0。这里的参考帧列表0对应的参考帧索引，参考帧列表1对应的参考帧索引就可以理解为上述的参考帧的信息。在一些实施例中，用两个标志位来分别表示是否使用参考帧列表0对应的运动信息以及是否使用参考帧列表0对应的运动信息，分别记为predFlagL0和predFlagL1。也可以理解为predFlagL0和predFlagL1表示上述单向运动信息“是否有效”。虽然没有明确地提到运动信息这种数据结构，但是它用每个参考帧列表对应的参考帧索引，运动矢量以及“是否有效”的标志位一起来表示运动信息。在一些标准文本中不出现运动信息，而是使用的运动矢量，也可以认为参考帧索引和是否使用对应运动信息的标志是运动矢量的附属。本申请中为了描述方便仍然用“运动信息”，但是应当理解，也可以用“运动矢量”来描述。
当前块所使用的运动信息可以保存下来。当前帧的后续编解码的块可以根据相邻的位置关系使用前面已编解码的块，如相邻块，的运动信息。这利用了空域上的相关性，所以这种已编解码的运动信息叫做空域上的运动信息。当前帧的每个块所使用的运动信息可以保存下来。后续编解码的帧可以根据参考关系使用前面已编解码的帧的运动信息。这利用了时域上的相关性，所以这种已编解码的帧的运动信息叫做时域上的运动信息。当前帧的每个块所使用的运动 信息的存储方法通常将一个固定大小的矩阵，如4x4的矩阵，作为一个最小单元，每个最小单元单独存储一组运动信息。这样每编解码一个块，它的位置对应的那些最小单元就可以把这个块的运动信息存储下来。这样使用空域上的运动信息或时域上的运动信息时可以直接根据位置找到该位置对应的运动信息。如一个16x16的块使用了传统的单向预测，那么这个块对应的所有的4x4个最小单元都存储这个单向预测的运动信息。如果一个块使用了GPM或AWP，那么这个块对应的所有的最小单元会根据GPM或AWP的模式，第一个运动信息，和第二个运动信息以及每个最小单元的位置确定每个最小单元存储的运动信息。一种方法是如果一个最小单元对应的4x4的像素全部来自于第一个运动信息，那么这个最小单元存储第一个运动信息，如果一个最小单元对应的4x4的像素全部来自于第二个运动信息，那么这个最小单元存储第二个运动信息。如果一个最小单元对应的4x4的像素既来自于第一个运动信息又来自于第二个运动信息，那么AWP会选择其中一个运动信息进行存储；GPM的做法是如果两个运动信息指向不同的参考帧列表，那么把它们组合成双向运动信息存储，否则只存储第二个运动信息。
可选的，上述mergeCandList是根据空域运动信息，时域运动信息，基于历史的运动信息，还有一些其他的运动信息来构建的。示例性的，mergeCandList使用如图6A中1至5的位置来推导空域运动信息，使用如图6A中的6或7的位置来推导时域运动信息。基于历史的运动信息是在每编解码一个块时，把这个块的运动信息添加到一个先进先出的列表里，添加过程可能需要一些检查，如是否跟列表里现有的运动信息重复。这样在编解码当前块时就可以参考这个基于历史的列表里的运动信息。
在一些实施例中，关于GPM的语法描述如表1所示：
表1
Figure PCTCN2021143977-appb-000001
如表1所示，在merge模式下，如果regular_merge_flag不为1，当前块可能用CIIP或GPM。如果当前块不用CIIP，那么它就用GPM，也就是在表1中的语法“if(！ciip_flag[x0][y0])”所示的内容。
如上述表1可知，GPM需要在码流中传输3个信息，即merge_gpm_partition_idx，merge_gpm_idx0，merge_gpm_idx1。x0，y0用来确定当前块左上角亮度像素相对于图像左上角亮度像素的坐标(x0,y0)。merge_gpm_partition_idx确定GPM的划分形状，如上所示，它是“模拟划分”，merge_gpm_partition_idx为本申请实施例所说的权重导出模式或者说权重导出模式的索引。merge_gpm_idx0是第一个运动信息在候选列表中的索引值，merge_gpm_idx1是第二个运动信息在候选列表中的索引值。如果候选列表长度(MaxNumGpmMergeCand)>2才需要传merge_gpm_idx1，否则可以直接确定。
在一些实施例中，GPM的解码过程包括如下步骤：
解码过程输入的信息包括：当前块左上角的亮度位置相对于图像左上角的坐标(xCb,yCb)，当前块亮度分量的宽度cbWidth，当前块亮度分量的高度cbHeight，1/16像素精度的亮度运动矢量mvA和mvB，色度运动矢量mvCA和mvCB，参考帧索引refIdxA和refIdxB，预测列表标志predListFlagA和predListFlagB。
示例性的，可以用运动矢量，参考帧索引和预测列表标志组合起来来表示运动信息。在一些实施例中，支持2个参考帧列表，每个参考帧列表可能有多个参考帧。单向预测只使用其中一个参考帧列表中的一个参考帧的一个参考块作为参考，双向预测使用两个参考帧列表中的各一个参考帧的各一个参考块作为参考。而GPM使用2各单向预测。 上述mvA和mvB，mvCA和mvCB，refIdxA和refIdxB，predListFlagA和predListFlagB中的A可以理解为第一预测模式，B可以理解为第二预测模式。可选的，用X表示A或B，predListFlagX表示X使用第一个参考帧列表还是第二个参考帧列表，refIdxX表示X使用的参考帧列表中的参考帧索引，mvX表示X使用的亮度运动矢量，mvCX表示X使用的色度运动矢量。需要说明的是，可以认为使用运动矢量，参考帧索引和预测列表标志组合起来来表示本申请所述的运动信息。
解码过程输出的信息包括：(cbWidth)X(cbHeight)的亮度预测样本矩阵predSamplesL；(cbWidth/SubWidthC)X(cbHeight/SubHeightC)的Cb色度分量的预测样本矩阵，如果需要；(cbWidth/SubWidthC)X(cbHeight/SubHeightC)的Cr色度分量的预测样本矩阵，如果需要。
示例性的，下面以亮度分量举例，色度分量的处理和亮度分量类似。
假设predSamplesLAL和predSamplesLBL的大小为(cbWidth)X(cbHeight)，为根据2个预测模式做出的预测样本矩阵。predSamplesL按如下方法导出：分别根据亮度运动矢量mvA和mvB，色度运动矢量mvCA和mvCB，参考帧索引refIdxA和refIdxB，预测列表标志predListFlagA和predListFlagB确定predSamplesLAL和predSamplesLBL。即分别根据2个预测模式的运动信息进行预测，详细过程不再赘述。通常GPM是merge模式，可以认为GPM的2个预测模式都是merge模式。
根据merge_gpm_partition_idx[xCb][yCb]，利用表2确定GPM的“划分”角度索引变量angleIdx和距离索引变量distanceIdx。
表2–angleIdx和distanceIdx与merge_gpm_partition_idx的对应关系
Figure PCTCN2021143977-appb-000002
需要说明的是，因为三个分量(component，如Y、Cb、Cr)都可以使用GPM，所以在一些标准文本将一个分量产生GPM的预测样本矩阵的过程分装到了一个子流程里面，即GPM的加权预测过程(Weighted sample prediction process for geometric partitioning mode)，三个分量都会调用这个流程，只是调用的参数不同，这里只用亮度分量举例。当前亮度块的预测矩阵predSamplesL[xL][yL](其中xL＝0..cbWidth–1，yL＝0..cbHeight-1)由GPM的加权预测过程导出。其中nCbW设为cbWidth,nCbH设为cbHeight，两个预测模式做的预测样本矩阵predSamplesLAL和predSamplesLBL,还有angleIdx,distanceIdx作为输入。
在一些实施例中，GPM的加权预测导出过程包括如下步骤：
该过程的输入有：当前块的宽度nCbW，当前块的高度nCbH；2个(nCbW)X(nCbH)的预测样本矩阵predSamplesLA和predSamplesLB；GPM的“划分”角度索引变量angleIdx；GPM的距离索引变量distanceIdx；分量索引变量cIdx。本示例以亮度举例，因此上述cIdx为0，表示亮度分量。
该过程的输出有：(nCbW)X(nCbH)的GPM预测样本矩阵pbSamples。
示例性的，变量nW,nH,shift1,offset1,displacementX,displacementY,partFlip还有shiftHor按如下方法导出:
nW＝(cIdx＝＝0)？nCbW:nCbW*SubWidthC；
nH＝(cIdx＝＝0)？nCbH:nCbH*SubHeightC；
shift1＝Max(5,17-BitDepth)，其中BitDepth是编解码的比特深度；
offset1＝1<<(shift1-1)，其中“<<”表示左移；
displacementX＝angleIdx；
displacementY＝(angleIdx+8)％32；
partFlip＝(angleIdx>＝13&&angleIdx<＝27)？0:1；
shiftHor＝(angleIdx％16＝＝8||(angleIdx％16！＝0&&nH>＝nW))？0:1。
变量offsetX和offsetY按如下方法导出:
当shiftHor的值为0时:
offsetX＝(-nW)>>1，
offsetY＝((-nH)>>1)+(angleIdx<16？(distanceIdx*nH)>>3:-((distanceIdx*nH)>>3))。
当shiftHor的值为1时:
offsetX＝((-nW)>>1)+(angleIdx<16？(distanceIdx*nW)>>3:-((distanceIdx*nW)>>3)，
offsetY＝(-nH)>>1。
变量xL和yL按如下方法导出:
xL＝(cIdx＝＝0)？x:x*SubWidthC，
yL＝(cIdx＝＝0)？y:y*SubHeightC，
表示当前位置预测样本权重的变量wValue按如下方法导出，wValue即为(x，y)点的第一预测模式的预测矩阵的预测值predSamplesLA[x][y]的权重，而(8-wValue)即为(x，y)点的第一预测模式的预测矩阵的预测值predSamplesLB[x][y]的权重。
其中距离矩阵disLut按表3确定:
表3
idx 0 2 3 4 5 6 8 10 11 12 13 14
disLut[idx] 8 8 8 4 4 2 0 -2 -4 -4 -8 -8
idx 16 18 19 20 21 22 24 26 27 28 29 30
disLut[idx] -8 -8 -8 -4 -4 -2 0 2 4 4 8 8
weightIdx＝(((xL+offsetX)<<1)+1)*disLut[displacementX]+(((yL+offsetY)<<1)+1)*disLut[displacementY]，
weightIdxL＝partFlip？32+weightIdx:32–weightIdx，
wValue＝Clip3(0,8,(weightIdxL+4)>>3)，
预测样本的值pbSamples[x][y]按如下方法导出:
pbSamples[x][y]＝Clip3(0,(1<<BitDepth)-1,(predSamplesLA[x][y]*wValue+predSamplesLB[x][y]*(8-wValue)+offset1)>>shift1)。
需要说明的是，对当前块的每一个位置推导一个权重值，然后计算一个GPM的预测值pbSamples[x][y]。因为这种方式权重wValue不必写成一个矩阵的形式，但是可以理解如果把每个位置的wValue都保存到一个矩阵里，那它就是一个权重矩阵。每个点分别计算权重并加权得到GPM的预测值，或者计算出所有的权重再统一加权得到GPM的预测样本矩阵其原理是一样的。而本申请中的诸多描述中使用权重矩阵的说法，是为了表述更容易理解，用权重矩阵画图更直观，其实也可以按每个位置的权重来描述。比如权重矩阵导出模式也可以说成权重导出模式。
在一些实施例中，如图6B所示，GPM的解码流程可以表述为：解析码流，确定当前块是否使用GPM技术；如果当前块使用GPM技术，确定权重导出模式(或“划分”模式或权重矩阵导出模式)，及第一运动信息和第二运动信息。分别根据第一运动信息确定第一预测块，根据第二运动信息确定第二预测块，根据权重矩阵导出模式确定权重矩阵，根据第一预测块和第二预测块和权重矩阵确定当前块的预测块。
需要说明的是，在本申请的实施例中，GPM或AWP属于一种预测技术，GPM或AWP需要在码流中传输一个GPM或AWP是否使用的标志(flag)，该flag可以指示当前块是否使用GPM或AWP。如果使用GPM或AWP，编码器在码流中需要传输具体使用的模式，即GPM的64种划分模式之一，或AWP的56种划分模式之一；以及两个单向运动信息的索引值。也就是说，对于当前块而言，解码器通过解析码流可以得到GPM或AWP是否使用的信息，如果确定使用GPM或AWP，解码器可以解析出GPM或AWP的预测模式参数以及两个运动信息索引值，比如当前块可以划分为两个分区，那么可以解析出第一分区对应的第一索引值和第二分区对应的第二索引值。
具体来讲，对于GPM模式来说，如果使用GPM，那么码流中将会传输GPM下的预测模式参数，比如GPM具体的划分模式；通常情况下，GPM包括有64种划分模式。对于AWP模式来说，如果使用AWP，那么码流中将会传输AWP下的预测模式参数，比如AWP具体的划分模式；通常情况下，AWP包括有56种划分模式。
在帧间预测模式下，比如GPM和AWP均需要使用两个单向运动信息查找两个参考块。目前的实现方式是在编码器侧利用当前块之前已编码/已解码部分的相关信息构建一个单向运动信息候选列表，从单向运动信息候选列表中选择单向运动信息，将这两个单向运动信息在单向运动信息候选列表中的索引值(index)写入码流。在解码器侧采用同样的方式，即利用当前块之前已解码部分的相关信息构建一个单向运动信息候选列表，这个单向运动信息候选列表与编码器侧构建的候选列表一定是相同的。如此，从码流中解析出两个单向运动信息的索引值，然后从单向运动信息候选列表中查找出这两个单向运动信息即为当前块需要使用的两个单向运动信息。
也就是说，本申请所描述的单向运动信息可以包括：运动矢量信息，即(x，y)的值，以及对应的参考帧信息，即参考帧列表及在参考帧列表中的参考帧索引值。一种表示方式是记录两个参考帧列表的参考帧索引值，其中一个参考帧列表对应的参考帧索引值有效，如0，1，2等；另一个参考帧列表对应的参考帧索引值为无效，即-1。参考帧索引值有效的参考帧列表即为当前块的运动信息所使用的参考帧列表，根据参考帧索引值可以从该参考帧列表中查找到对应的参考帧。每个参考帧列表都有一个对应的运动矢量，有效的参考帧列表对应的运动矢量是有效的，无效的参考帧列表对应的运动矢量是无效的。解码器可以通过单向运动信息中的参考帧信息找到所需的参考帧，根据当前块的位置以及运动矢量即(x，y)的值可以在参考帧中找到参考块，进而确定出当前块的帧间预测值。
帧内预测方法是使用当前块周边已编解码的重建像素作为参考像素来对当前块进行预测。图7A为帧内预测的示意图，如图7A所示，当前块的大小为4x4，当前块左边一行和上面一列的像素为当前块的参考像素，帧内预测使用这些参考像素对当前块进行预测。这些参考像素可能已经全部可得，即全部已经编解码。也可能有部分不可得，比如当前块是整帧的最左侧，那么当前块的左边的参考像素不可得。或者编解码当前块时，当前块左下方的部分还没有编解码，那么左下方的参考像素也不可得。对于参考像素不可得的情况，可以使用可得的参考像素或某些值或某些方法进行填充，或者不进行填充。
图7B为帧内预测的示意图，如图7B所示，多参考行帧内预测方法(Multiple reference line，MRL)可以使用更多的参考像素从而提高编解码效率，例如，使用4个参考行/列为当前块的参考像素。
进一步地，帧内预测有多种预测模式，图8A-5I为帧内预测的示意图，如图8A-5I所示，H.264中对4x4的块进行帧内预测主要可以包括9种模式。其中，如图8A所示的模式0将当前块上面的像素按垂直方向复制到当前块作为预 测值，如图8B所示的模式1将左边的参考像素按水平方向复制到当前块作为预测值，如图8C所示的模式2直流DC将A～D和I～L这8个点的平均值作为所有点的预测值，如图8D-5I所示的模式3～8分别按某一个角度将参考像素复制到当前块的对应位置，因为当前块某些位置不能正好对应到参考像素，可能需要使用参考像素的加权平均值，或者说是插值的参考像素的分像素。
除此之外，还有Planar模式等，而随着技术的发展以及块的扩大，角度预测模式也越来越多。图9为帧内预测模式的示意图，如图9所示，如HEVC使用的帧内预测模式有Planar、DC和33种角度模式共35种预测模式。图10为帧内预测模式的示意图，如图10所示，VVC使用的帧内模式有Planar、DC和65种角度模式共67种预测模式。图11为帧内预测模式的示意图，如图11所示，AVS3使用DC、Planar、Bilinear和63种角度模式共66种预测模式。
另外还有一些技术对预测进行改进，如改进参考像素的分像素插值，对预测像素进行滤波等。如AVS3中的多组合帧内预测滤波(multipleintraprediction filter，MIPF)对不同的块大小，使用不同的滤波器产生预测值。对同一个块内的不同位置的像素，与参考像素较近的像素使用一种滤波器产生预测值，与参考像素较远的像素使用另一种滤波器产生预测值。对预测像素进行滤波的技术如AVS3中的帧内预测滤波(intraprediction filter，IPF)，对预测值可以使用参考像素进行滤波。
在帧内预测中可以使用最可能模式列表(MostprobableModes List，MPM)的帧内模式编码技术来提高编解码效率。利用周边已编解码的块的帧内预测模式，以及根据周边已编解码块的帧内预测模式导出的帧内预测模式，如相邻的模式，以及一些常用或使用概率比较高的帧内预测模式，如DC，Planar，Bilinear模式等，构成一个模式列表。参考周边已编解码的块的帧内预测模式利用了空间上的相关性。因为纹理在空间上会有一定的连续性。MPM可以作为帧内预测模式的预测。也就是认为当前块使用MPM的概率会比不使用MPM的概率高。因而在二值化时，会给MPM使用更少的码字，从而节省开销，以提高编解码效率。
GPM用权重矩阵组合两个帧间预测块。实际上它可以扩展到组合两个任意的预测块。如两个帧间预测块，两个帧内预测块，一个帧间预测块和一个帧内预测块。甚至在屏幕内容编码中，还可以使用IBC(intra block copy)或palette的预测块作为其中的1个或2个预测块。
本申请将帧内、帧间、IBC、palette称为不同的预测方式。为了表述方便，这里使用一个叫预测模式的称呼。预测模式可以理解为根据它编解码器可以产生当前块的一个预测块的信息。比如，在帧内预测中，预测模式可以是某个帧内预测模式，如DC，Planar，各种帧内角度预测模式等。当然也可以叠加某个或某些辅助的信息，比如帧内参考像素的优化方法，产生初步的预测块以后的优化方法(比如滤波)等。比如，在帧间预测中，预测模式可以是skip(跳过)模式，merge(合并)模式或MMVD(merge with motion vector difference，带运动矢量差的合并)模式，或普通的inter模式(MVP+MVD)，可以是单向预测也可以是双向预测或多假设预测。如果帧间的预测模式使用单向预测，它要能确定一个运动信息，这个运动信息是一个单向运动信息，根据运动信息能确定预测块。如果帧间的预测模式使用双向预测，它要能确定一个双向运动信息或两个单向运动信息，根据运动信息能确定预测块。如果帧间的预测模式使用多假设预测，它要能确定多个单向运动信息，根据运动信息能确定预测块。skip，merge，普通的inter模式都可以支持单向预测，双向预测或多假设预测。一个预测模式如果是帧间预测模式，它能确定运动信息，根据运动信息能确定预测块。在skip模式和merge模式，MMVD模式，普通inter模式的基础上都可以使用模板匹配的方法，这样的预测模式可以仍然称为skip模式和merge模式，MMVD模式，普通inter模式或者使用模板匹配的skip模式，使用模板匹配的merge模式，使用模板匹配的MMVD模式，使用模板匹配的普通inter模式。
skip模式和merge模式都不需要在码流中传输运动矢量差MVD，skip模式还不需要在码流种传输残差。而MMVD可以认为是一种特殊的merge模式，它通过一些标志位来表示一些特定的MVD，这些特定的MVD只有几种可能的预设值。一个例子是VVC中的MMVD模式，它用mmvd_direction_idx表示MVD的方向，mmvd_direction_idx可能的值是0，1，2，3。0表示MMVD的水平分量为正值，竖直方向为0；1表示MMVD的水平分量为负值，竖直方向为0；2表示MMVD的水平分量为0，竖直方向为正值；3表示MMVD的水平分量为0，竖直方向为负值。用mmvd_distance_idx表示上述正值或负值的绝对值，mmvd_distance_idx可能的值是0～7，在ph_mmvd_fullpel_only_flag＝＝0时分别表示1，2，4，8，16，32，64，128，在ph_mmvd_fullpel_only_flag＝＝1时分别表示4，8，16，32，64，128，256，512。而普通的inter模式的MVD理论上可以表示一个有效范围内的任意可能的MVD。
这样GPM需要确定的信息可以表述为1个权重导出模式和2个预测模式。权重导出模式用来确定权重矩阵或权重，2个预测模式分别确定一个预测块或预测值。权重导出模式在某些地方也被称为划分模式。但因为它是模拟划分，本申请更习惯称为权重导出模式。
可选的，2个预测模式可以来自相同的或不同的预测方式，其中预测方式包括但不限于帧内预测、帧间预测、IBC、palette。
一个具体的具体的例子如下：如果当前块使用GPM。这个例子用在帧间编码的块中，允许使用帧内预测和帧间预测中的merge模式。如表4所示，增加一个语法元素intra_mode_idx表示哪一个预测模式是帧内预测模式，比如intra_mode_idx为0表示2个预测模式都是帧间预测模式，即mode0IsInter为1，mode0IsInter为1；intra_mode_idx为1表示第一预测模式是帧内预测模式，第二预测模式是帧间预测模式，即mode0IsInter为0，mode0IsInter为1；intra_mode_idx为2表示第一预测模式是帧间预测模式，第二预测模式是帧内预测模式，即mode0IsInter为1，mode0IsInter为0；intra_mode_idx为3表示两个预测模式都是帧内预测模式，即mode0IsInter为0，mode0IsInter为0。
表4
Figure PCTCN2021143977-appb-000003
Figure PCTCN2021143977-appb-000004
在一些实施例中，如图12所示，GPM的解码流程可以表述为：解析码流，确定当前块是否使用GPM技术；如果当前块使用GPM技术，确定权重导出模式(或“划分”模式或权重矩阵导出模式)，及第一帧内预测模式和第二帧内预测模式。分别根据第一帧内预测模式确定第一预测块，根据第二帧内预测模式确定第二预测块，根据权重矩阵导出模式确定权重矩阵，根据第一预测块和第二预测块和权重矩阵确定当前块的预测块。
模板匹配(template matching)的方法最早用在帧间预测中，它利用相邻像素之间的相关性，把当前块周边的一些区域作为模板。在当前块进行编解码时，按照编码顺序其左侧及上侧已经编解码完成。当然在现有的硬件解码器实现时，不一定能保证当前块开始解码时，其左侧和上侧已经解码完成，当然这里说的是帧间块，比如在HEVC中帧间编码的块产生预测块时是不需要周边的重建像素的，因而帧间块的预测过程可以并行进行。但是帧内编码的块是一定需要左侧和上侧的重建像素作为参考像素的。理论上左侧和上侧是可得的，也就是说硬件设计做相应的调整是可以实现的。相对来说右侧和下侧在现在标准如VVC的编码顺序下是不可得的。
如图13所示把当前块的左侧和上侧的矩形区域设为模板，左侧的模板部分的高度一般和当前块的高度相同，上侧的模板的部分的宽度一般和当前块的宽度相同，当然也可以不同。在参考帧中寻找模板的最佳匹配位置从而确定当前块的运动信息或者说运动矢量。这个过程大致可以描述为，在某一个参考帧中，从一个起始位置开始，在周边一定范围内进行搜索。可以预先设定好搜索的规则，如搜索范围搜索步长等。每移动一个到位置，计算该位置对应的模板和当前块周边的模板的匹配程度，所谓匹配程度可以用一些失真代价来衡量，比如说SAD(sum of absolute difference)，SATD(sum of absolute transformed difference)，一般SATD使用的变换是Hadamard变换，MSE(mean-square error)等，SAD，SATD，MSE等的值越小代表匹配程度越高。用该位置对应的模板的预测块和当前块周边的模板的重建块计算代价。除了整像素位置的搜索还可以进行分像素位置的搜索，根据搜索到的匹配程度最高的位置来确定当前块的运动信息。利用相邻像素之间的相关性，对模板合适的运动信息可能也是当前块合适的运动信息。当然模板匹配的方法可能并不一定对所有的块都适用，因而可以使用一些方法确定当前块是否使用上述模板匹配的方法，比如在当前块用一个控制开关表示是否使用模板匹配的方法。这种模板匹配的方法的一个名字叫DMVD(decoder side motion vector derivation)。编码器和解码器都可以利用模板进行搜索从而导出运动信息或者在原有的运动信息的基础上找到更好的运动信息。而它不需要传输具体的运动矢量或运动矢量差，而是由编码器和解码器都进行同样规则的搜索从而保证编码和解码的一致。模板匹配的方法可以提高压缩性能，但是它需要在解码器中也进行“搜索”，从而带来了一定的解码器复杂度。
上述是在帧间上应用模板匹配的方法，模板匹配的方法也可以用在帧内上，比如说利用模板来确定帧内预测模式。对当前块，同样可以使用当前块上边和左边一定范围内的区域作为模板，比如说仍然如上图所示的左边的矩形区域和上边的矩形区域。在编解码当前块时，在模板中的重建像素是可得的。这个过程大致可以描述为，对当前块确定候选的帧内预测模式的集合，候选的帧内预测模式构成全部可用的帧内预测模式的一个子集。当然候选的帧内预测模式可以是全部可用的帧内预测模式的全集。这可以根据性能和复杂度的权衡来确定。可以根据MPM或一些规则，如等间距筛选等，来确定候选的帧内预测模式的集合。计算各候选的帧内预测模式在模板上的代价，比如SAD，SATD，MSE等。用该模式在模板上进行预测做出预测块，用预测块和模板的重建块计算代价。代价小的模式可能与模板更匹配，利用相邻像素之间的相似性，在模板上表现好的帧内预测模式可能也是当前块上表现好的帧内预测模式。选定1个或几个代价小的模式。当然上述2步可以重复进行，比如说在选定1个或几个代价小的模式后，再一次确定候选的帧内预测模式的集合，对新确定的候选的帧内预测模式集合再计算代价，选定1个或几个代价小的模式。这也可以理解为粗选和细选。最终选定的1个帧内预测模式确定为当前块的帧内预测模式，或者最终选定的几个帧内预测模式作为当前块的帧内预测模式的候选。当然也可以仅仅拿模板匹配的方法对候选的帧内预测模式集合进行排序，比如对MPM列表进行排序，即将MPM列表中的模式分别在模板上做出预测块并确定代价，按代价从小到大进行排序。一般MPM列表中越靠前的模式在码流中的开销越小，这样也可以达到提高压缩效率的目的。
模板匹配的方法可以用于确定GPM的2个预测模式上。如果将模板匹配的方法用于GPM，对当前块可以用1个控制开关控制当前块的2个预测模式是否使用模板匹配，也可以用2个控制开关分别控制2个预测模式各自是否使用模板匹配。
另一方面是如何使用模板匹配。比如如果GPM在merge模式下使用，如VVC中的GPM，它使用merge_gpm_idxX从mergeCandList中确定一个运动信息，其中大写的X为0或1。对第X个运动信息，一种方法是在上述运动信息的基础上用模板匹配的方法进行优化。即根据merge_gpm_idxX从mergeCandList中确定一个运动信息，如果对该运动信息使用模板匹配，那么用模板匹配的方法在上述运动信息的基础上进行优化。另一种方法是不使用merge_gpm_idxX从mergeCandList中确定一个运动信息，而是直接从一个默认运动信息的基础上进行搜索，确定一个运动信息。
如果第X预测模式是帧内预测模式，而且当前块的第X预测模式使用模板匹配的方法，那么可以利用模板匹配方法确定一个帧内预测模式，不需要在码流中指示该帧内预测模式的索引。或者利用模板匹配方法确定一个候选集合或者MPM列表，需要在码流中指示该帧内预测模式的索引。
GPM确定了权重导出模式后，就可以确定每个预测模式占领的区域。所谓占领的区域可以理解为该预测模式对应的权重为最大值的区域，或者权重大于或等于某一阈值的区域。GPM之所以能提高压缩性能，就是因为GPM“划分”的2个部分是不同的。所以当用模板匹配的方法来确定GPM的预测模式时，也可以对模板进行划分。现有技术可以将模板分为3类，即左侧，上侧和全部(左侧加上侧)。模板的划分是跟权重导出模式有关的。示例性的，如表5所示，现有技术里模板的划分跟“划分”角度或者说“划分”角度索引angleIdx有关。
表5
“划分”角度索引 第一预测模式对应的模板 第二预测模式对应的模板
0 TM_A TM_AL
1 / /
2 TM_A TM_AL
3 TM_A TM_AL
4 TM_A TM_L
5 TM_AL TM_L
6 / /
7 / /
8 TM_AL TM_L
9 / /
10 / /
11 TM_AL TM_L
12 TM_AL TM_AL
13 TM_A TM_AL
14 TM_A TM_AL
15 / /
16 TM_A TM_AL
17 / /
18 TM_A TM_AL
19 TM_A TM_AL
20 TM_A TM_L
21 TM_AL TM_L
22 / /
23 / /
24 TM_AL TM_L
25 / /
26 / /
27 TM_AL TM_L
28 TM_AL TM_AL
29 TM_A TM_AL
30 TM_A TM_AL
31 / /
例如，记左侧模板为TM_A，上侧模板为TM_L，全部(左侧加上侧)模板为TM_AL。模板和“划分”角度索引的关系如表5所示，其中某些角度索引如1，6，7等在目前的GPM中并没有用到，所以没有对应的模板，用/表示。
根据权重导出模式划分模板确实考虑到了GPM“划分”的2个部分的区别，但是这样的划分其实并不够精细，因为它只是将模板分为左侧和上侧2个部分，而根据权重图可以看到分界线(分界线可以认为是权重矩阵中权重为中值的点组成的线，在现在的GPM中划分线是直线，如果实际权重矩阵中没有权重为中值的整像素点，那么可以用分像素点来代替，当然也可以使用某一其他权重的点)可能落在各种位置。这样根据上述表5选择模板时，可能会选择不合适的模块，造成预测模式匹配不准确，进而导致预测精度低，编码效果差的问题。
为了解决上述技术问题，在本申请的实施例中，根据当前块的大小和权重导出模式中的至少一个，为当前块确定K个模板，使用这K个模板确定K个预测模式。也就是说本申请在确定K个模板时是基于当前块的大小和/或权重导 出模式的，使得确定出的K个模板更加符合实际情况，这样使用这K个模板确定预测模式时，可以提高预测模式的确定准确性，进而使用准确确定的K个预测模式实现当前块的准确预测，提升编码效果。
下面结合图14，以解码端为例，对本申请实施例提供的视频解码方法进行介绍。
图14为本申请一实施例提供的预测方法流程示意图，本申请实施例应用于图1和图3所示视频解码器。如图14所示，本申请实施例的方法包括：
S101、解码码流，确定当前块的权重导出模式。
需要说明的是，在本申请中，权重导出模式用于对当前块使用的权重进行确定。具体地，权重导出模式可以是导出权重的模式。对于一个给定长度和宽度的块，每一种权重导出模式可以导出一个权重矩阵；对于同样大小的块，不同权重导出模式导出的权重矩阵不同。
示例性的，在本申请中，AWP有56种权重导出模式，GPM有64种权重导出模式。
本申请中，解码端确定当前块的权重导出模式的方式包括但不限于如下几种：
方式一，解码端默认选取与编码端相同的权重导出模式，例如，解码端和编码端均选择索引为44的权重导出模式。
方式二，编码端将在编码过程中使用的权重导出模式的索引携带在码流中。这样解码端解码码流，可以得到当前块的权重导出模式。
方式三，采用与编码端相同的方式，确定出权重导出模式。例如，解码端尝试K个预测模式和权重导出模式的所有可能组合，K为大于1的正整数，选出代价最小的一个组合中的权重导出模式，确定为当前块的权重导出模式。
以K等于2为例，上述K个预测模式包括第一预测模式和第二预测模式，假设所有可用的预测模式有66种，第一预测模式有66种可能，由于第二预测模式与第一预测模式不相同，因此第二预测模式有65种，假设权重导出模式有64种(以GPM为例)，那么本申请可能使用任意两种不同的预测模式以及任意一种权重导出模式，共有66×65×64种可能。如果设定不使用PCM这种预测模式。那么就有65×64×63种可能。可见，在本申请中，还可以限制可以选择的预测模式，以及限制可以使用的权重导出模式的个数，那么组合的情况也会相应减少。
进一步地，在本申请的实施例中，解码器可以对所有可能组合进行代价计算，确定代价最小的一个组合。
假设K＝2时，则每一个组合均为包括有一个第一预测模式、一个第二预测模式和一个权重导出模式。
可选地，为了减小代价计算的耗时，可以先对上述所有可能组合进行初选，如使用SAD，和SATD等作为近似的代价进行初选，确定设定数量的候选第一预测模式、第二预测模式、权重导出模式的组合，再进行更详细的代价计算以实现细选，确定代价最小的一个第一预测模式、第二预测模式、权重导出模式的组合。从而可以在初选时使用一些快速算法减少尝试的次数，比如说一个角度预测模式造成代价很大时，与它相邻的几个预测模式都不再尝试等。
可以理解的是，在本申请中，上述初选和细选时都会根据第一预测模式确定第一预测值，根据第二预测模式确定第二预测值，根据权重导出模式导出权重，根据第一预测值、第二预测值和权重确定本申请的预测值。SAD和SATD初选时使用当前块和当前块对应的预测值来确定SAD和SATD。需要说明的是，上述根据权重导出模式导出的权重可以理解为导出当前块中每一个像素点对应的权重，也可以理解为导出当前块对应的权重矩阵。其中，基于权重确定当前块的预测值时，可以是确定当前块中每一个像素点对应的第一预测值和第二预测值，并根据每一个像素点对应的第一预测值、第二预测值和权重，确定出每个像素点对应的预测值，当前块中每一个像素点对应的预测值构成当前块的预测值。可选的，基于权重确定当前块的预测值时，还可以是按照块来执行，例如，确定当前块的第一预测值和第二预测值，根据当前块的权重矩阵对当前块的第一预测值和第二预测值进行加权，得到当前块的新的预测值。
在一些实施例中，解码端在确定当前块的权重导出模式之前，首先需要判断当前块是否使用K个不同的预测模式进行加权预测处理。若解码端确定当前块使用K个不同的预测模式进行加权预测处理时，则执行上述S101确定当前块的权重导出模式。
在一种可能的实现方式中，解码端可以通过确定当前块的预测模式参数，来确定当前块是否使用K个不同的预测模式进行加权预测处理。
可选的，在本申请的实施中，预测模式参数可以指示当前块是否可以使用GPM模式或AWP模式，即指示当前块是否可以使用K个不同的预测模式进行预测处理。
可以理解的是，在本申请的实施例中，可以将预测模式参数理解为一个表明是否使用了GPM模式或AWP模式标志位。具体地，编码器可以使用一个变量作为预测模式参数，从而可以通过对该变量的取值的设置来实现预测模式参数的设置。示例性的，在本申请中，如果当前块使用GPM模式或AWP模式，那么编码器可以将预测模式参数的取值设置为指示当前块使用GPM模式或AWP模式，具体地，编码器可以将变量的取值设置为1。示例性的，在本申请中，如果当前块不使用GPM模式或AWP模式，那么编码器可以将预测模式参数的取值设置为指示当前块不使用GPM模式或AWP模式，具体地，编码器可以将变量取值设置为0。进一步地，在本申请的实施例中，编码器在完成对预测模式参数的设置之后，便可以将预测模式参数写入码流中，传输至解码器，从而可以使解码器在解析码流之后获得预测模式参数。
基于此，解码端解码码流，得到预测模式参数，进而根据该预测模式参数确定当前块是否使用GPM模式或AWP模式，若当前块使用GPM模式或AWP模式，即使用K个不同的预测模式进行预测处理时，确定当前块的权重导出模式。
需要说明的是，在本申请的实施例中，GPM模式或AWP模式为一种预测方法，具体地，为当前块确定K个不同的预测模式，然后分别根据这K个不同的预测模式确定出K个预测值，接着可以再确定权重，将则会K个预测值依据权重进行组合，最终便可以得到新的预测值。
上述当前块的K个不同的预测模式包括如下几种示例：
示例1，上述K个不同的预测模式均为帧内预测模式。
示例2，上述K个不同的预测模式均为帧间预测模式。
示例3，上述K个不同的预测模式中，至少一个为帧内预测模式，至少一个为帧间预测模式。
示例4，上述K个不同的预测模式中，至少一个为帧内预测模式，至少一个为非帧间和非帧内的预测模式，例如为块内复制IBC预测模式或调色板palette预测模式等。
示例5，上述K个不同的预测模式中，至少一个为帧间预测模式，至少一个为非帧间和非帧内的预测模式，例如为IBC预测模式或palette预测模式等。
示例6，上述K个不同的预测模式均不是帧内预测模式，也不是帧间预测模式，例如一个为IBC预测模式，一个为palette预测模式等。
需要说明的是，本申请实施例对上述K个不同的预测模式的具体类型不做限制。
图15为使用两种预测模式对当前块进行预测时的示意图，如图15所示，在对当前块进行预测时，可以使用第一预测模式确定第一预测值，同时使用第二预测模式确定第二预测值，然后可以利用权重对第一预测值和第二预测值进行组合处理，最终获得一个新的预测值。
在一些实施例中，在应用GPM模式或AWP模式时，可以对当前块的尺寸进行限制。
可以理解的是，由于本申请实施例提出的预测方法需要分别使用K个不同的预测模式生成K个预测值，再根据权重进行加权得到新的预测值，为了降低的复杂度，同时考虑压缩性能和复杂度的权衡，在本申请的实施例中，可以限制对一些大小的块不使用该GPM模式或AWP模式。因此，在本申请中，解码器可以先确定当前块的尺寸参数，然后根据尺寸参数确定当前块是否使用GPM模式或AWP模式。
需要说明的是，在本申请的实施例中，当前块的尺寸参数可以包括当前块的高度和宽度，因此，解码器可以利用当前块的高度和宽度对使用GPM模式或AWP模式进行限制。
示例性的，在本申请中，若宽度大于第一阈值且高度大于第二阈值，则确定当前块使用GPM模式或AWP模式。可见，一种可能的限制是仅仅在块的宽度大于(或大于等于)第一阈值，且块的高度大于(或大于等于)第二阈值的情况下使用GPM模式或AWP模式。其中，第一阈值和第二阈值的值可以是8，16，32等，第一阈值可以等于第二阈值。
示例性的，在本申请中，若宽度小于第三阈值且高度大于第四阈值，则确定当前块使用GPM模式或AWP模式。可见，一种可能的限制是仅仅在块的宽度小于(或小于等于)第三阈值，且块的高度大于(或大于等于)第四阈值的情况下使用SAWP模式。其中，第三阈值和第四阈值的值可以是8，16，32等，第三阈值可以等于第四阈值。
进一步地，在本申请的实施例中，还可以通过像素参数的限制来实现限制能够使用GPM模式或AWP模式的块的尺寸。
示例性的，在本申请中，解码器可以先确定当前块的像素参数，然后再根据像素参数和第五阈值进一步判断当前块是否可以使用GPM模式或AWP模式。可见，一种可能的限制是仅仅在块的像素数大于(或大于等于)第五阈值的情况下使用GPM模式或AWP模式。其中，第五阈值的值可以是8，16，32等。
也就是说，在本申请中，只有在当前块的尺寸参数满足大小要求的条件下，当前块才可以使用GPM模式或AWP模式。
示例性的，在本申请中，可以有一个帧级的标志来确定当前待解码帧是否使用本申请。如可以配置帧内帧(如I帧)使用本申请，帧间帧(如B帧、P帧)不使用本申请。或者可以配置帧内帧不使用本申请，帧间帧使用本申请。或者可以配置某些帧间帧使用本申请，某些帧间帧不使用本申请。帧间帧也可以使用帧内预测，因而帧间帧也有可能使用本申请。
在一些实施例中，还可以有一个帧级以下、CU级以上(如tile、slice、patch、LCU等)的标志来确定这一区域是否使用本申请。
S102、根据当前块的大小和权重导出模式中的至少一个，确定K个模板。
其中，K为大于1的正整数。
模板匹配是利用相邻像素之间的相关性，把当前块周边的一些区域作为模板。在当前块进行编解码时，按照编码顺序其左侧及上侧已经解码完成。在帧间预测时，在参考帧中寻找模板的最佳匹配位置从而确定当前块的运动信息或者说运动矢量。在帧内预测时，利用模板来确定当前块的帧内预测模式。
本申请对当前块的模板的具体形状不做限制。
在一些实施例中，当前块的模板包括当前块的上方已解码区域和左侧已解码区域中的至少一个。
可选的，上方已解码区域的宽度与当前块的宽度相同，左侧已解码区域的高度与当前块的高度相同。
目前，如上述表5所示，第一预测模式和第二预测模式对应的模板为当前块的上方已解码区域，或者为当前块左侧已解码区域，或者为当前块的左侧已解码和上方已解码区域。这样在模板匹配时，使用第一预测模式对应的模板确定第一预测模式，使用第二预测模式对应的模板确定第二预测模式。例如图16所示，以GPM的索引为2的权重导出模式为例，当前块的权重矩阵的白色区域为第一预测模式的预测值对应的权重，黑色区域为第二预测模式的预测值对应的权重。如图16所示，第一预测模式对应的模板为当前块的上方已解码区域，第二预测模式对应的模板为当前块的左侧已解码区域，但是，与第二预测模式靠近的模板包括左侧区域外，还包括部分上方已解码区域，因此，现有技术对于模板的划分不够精细，进而导致基于不精细的模板确定预测模式时，存在预测模式确定不准确，预测误差大的问题。
为了解决上述技术问题，本申请实施例通过当前块的大小和权重导出模式中的至少一个，来实现模板的精细划分。下面结合以下情况1和情况2所提出的方法，对上述S102中根据当前块的大小和权重导出模式中的至少一个，确定K个模板的过程进行详细介绍。
情况1，本申请实施例可以通过权重导出模式实现模板的更精细划分，具体的，上述S102包括如下步骤：
S102-A、根据权重导出模式，将当前块的模板划分为K个模板。
由上述图16所示，与第二预测模式相关的模板不仅包括左侧区域，还包括上部区域中的靠左侧部分，与第一预测模式相关的模板包括上部区域中靠右侧部分。这样将当前块的模板中的左侧区域和上部区域的靠左侧部分作为第二预测模式的模板进行第二预测模式匹配时，可以提高第二预测模式的匹配准确性。将当前块的模板中的右上半部分作为第一预测模式的模板进行第一预测模式匹配时，可以提高第一预测模式的匹配准确性。由此可知，基于权重导出模式，可以实现模板的准确划分，进而提高预测模式的准确确定，提高解码效果。
上述S102-A中根据权重导出模式，将当前块的模板划分为K个模板的方式包括但不限于如下几种：
方式一，根据权重导出模式对应的权重矩阵的分界线，将当前块的模板划分成K个模板。
例如图17A所示，本申请将当前块的权重导出模式对应的权重矩阵的分界线向当前块的模板进行延伸，以将当前块的模板进行划分，假设K＝2，则可以将分界线右侧的模板记为第一模板，将分界线左侧的模板记为第二模板。第一模板对应第一预测模式，第二模板对应的第二预测模式，在模板匹配时，可以使用第一模板导出第一预测模式，使用第二模板导出第二预测模式，进而实现预测模式的准确确定，提升解码效果。
在一些实施例中，根据上述方法划分的第一模板和第二模板可能不是长方形，例如图17A所示，第一模板和第二模板具有斜边，对不规则的模板计算代价较复杂。
为了降低模板匹配的复杂度，在一些实施例中，可以把第一模板和第二模板都划分成矩形，具体的，将分界线向当前块的模板中进行延伸，得到在当前块的模板中的延伸线，使用该延伸线将当前块的模板划分成第一模板和第二模板，其中第一模板与第二模板中间的分界线与延伸线相交，或者与延伸线不相交。示例性的，如图17B所示，第一模板和第二模板的分界线经过延伸线的一个端点且与当前块的长度边垂直。示例性的，如图17C所示，第一模板和第二模板的分界线经过延伸线的中点且与当前块的长度边垂直。
在一些实施例中，若K大于2时，可以根据预设的划分方式，对根据权重导出模式划分后的模板再进行划分。假设K＝3，以图17B所示的根据权重导出模式划分后的模板为例，则可以将左侧模板划分成两部分，例如，将左侧模板的下半部分划分为第三模板，左侧模板的剩余上半部分和原有的上方模板的左半部分划分为第二模板，将上方模板的右半部分划分为第一模板，进而将当前块的模板划分成3个模板。
该方式一中，根据权重矩阵的分界线将当前块的模板划分成K个模板，该划分方式简单，且可以实现对模板的准确划分。
在一些实施例中，还可以根据如下方式二的方法将当前块的模板划分成K个模板。
方式二，上述S102-A包括如下S102-A1和S102-A2步骤：
S102-A1、将当前块的模板划分为M个子模板，M为大于或等于K的正整数；
S102-A2、根据权重导出模式，将M个子模板对应到K个模板中。
在该方式二中，首先将当前块的模板划分成多个子模板，例如划分成M个子模板，接着确定每个子模块对应到哪个模板，进而实现K个模板的划分。
本申请实施例对上述子模板的划分方式不做限制。
在方式二的一种可能实现方式1中，S102-A1包括：根据权重导出模式，将当前块的模板划分为M个子模板。
示例1，根据权重导出模式确定权重矩阵，将权重矩阵向当前块的模板延伸，例如向左向上延伸，将权重矩阵覆盖到当前块的模板上面。例如图17D所示，可以选择把当前块左上侧的小矩形区域加入到当前块的模板中，将当前块的模板和当前块拼起来构成一个矩形。当然也可以只使用左侧部分和上方部分作为当前块的模板。如图17D所示，当前块的模板包括当前块的左侧区域和上部区域，右下方的矩形区域为当前块。将当前块的权重矩阵向当前块的模板方向延伸，覆盖当前块的模板，这样可以根据当前块的模板被权重矩阵的覆盖情况，将当前块的模板划分成M个子模板。例如，将当前块的模板中权重位于a0至a1区间内的点划分为第一子模板，将权重位于a1至a2区间内的点划分为第二子模板，以此类推，将权重位于aM-1至aM区间内的点划分为第M子模板。
示例性的，将图17D中黑色模板划分为第一子模板，将上侧灰色模板划分为第二子模板，将上侧白色模板划分为第二子模板。
示例性的，将图17D中左侧的黑色模板划分为第一子模板，将上侧的黑色模板划分为第二子模板，将上侧灰色模板划分为第三子模板，将上侧白色模板划分为第四子模板。
本申请对上述M个子模板的具体形状不做限制。
在一些实施例中，为了降低后续模式匹配计算的复杂度，则上述示例1将M个子模板划分为矩形。
示例2，根据权重导出模式，确定权重的分界线，将分界线向当前块的模板中进行延长，以将当前块的模板划分为M个子模板。
具体的，根据权重导出模式确定权重的分界线，由上述实施例的描述可知，该分界线是权重导出模式导出的当前块中各点的权重组成的权重矩阵中权重发生变化的点构成的直线(或曲线)，例如图17E中的斜线。将该分界线向当前块的模板中进行延长，将当前块的上方模板划分成两部分。这样可以根据权重分界线划分后的模板，确定M个子模板。例如，将图17E所示，将当前块的上方模板中，位于分界线右侧的划分为第一子模板，将位于分界线左侧的划分为第二子模板，将当前块的左侧模板划分为第三子模板，此时将当前块的模板划分为3个子模板。再例如，如图17F所示，还可以将当前块的左侧模板划分为多个子模板，例如划分为两个子模块，此时，将当前块的模板划分为4个子模板。在一些实施例中，还可以根据其他的规则，对分界线划分后的模板进行再划分，得到M个子模板。
在一些实施例中，根据上述方法划分的第一模板和第二模板可能不是长方形，例如图17E和图17F所示，第一子模板和第二子模板具有斜边，对不规则的模板进行模板匹配计算代价时较复杂。
为了降低模板匹配的计算复杂度，在一些实施例中，将分界线向当前块的模板中进行延长，得到分界线在当前块的模板的延伸线；使用延伸线，将当前块的模板划分为M个矩形子模板。例如图17G所示，使用延伸线将第一子模 板和第二子模板划分为矩形。图17G示出的第一子模板和第二子模板的分界线经过延伸线的左侧端点且与当前块的长度边垂直，可选的，第一子模板和第二子模板的分界线经过延伸线的右侧端点且与当前块的长度边垂直，或者第一子模板与第二子模板的分界线经过延伸线的中点且与当前块的长度边垂直。或者，第一子模板和第二子模板的分界线与延伸线不相交，且与当前块的长度边垂直。
该方式二中除了上述根据权重导出模式，将当前块的模板划分为M个子模板外，还可以采用下面实现方式2来将当前块的模板划分为M个子模板，具体如下所示。
在方式二的一种可能实现方式2中，根据预设的规则，将当前块的模板划分为M个子模板，即上述S102-A1包括如下步骤：
S102-A11、将当前块的上方模板划分成P个子模板；和/或，
S102-A12、将当前块的左侧模板划分成Q个子模板；
其中，P和Q均为小于或等于M的整数，且P与Q之和等于M。
本申请实施例中，当前块的模板包括当前块上方已解码的若干行像素行和当前块左侧已解码的若干列像素列，为了便于描述，本申请实施例将当前块的上方已解码的若干像素行记为当前块的上方模板，将当前块左侧已解码的若干列像素列记为当前块的左侧模板。在一些实施例中，当前块的模板还包括当前块的左上角已解码的区域，和/或包括当前块左下方的已解码区域等，本申请实施例对当前块的具体模板不做限制。本申请实施例主要对当前块的模板中的上方模板和左侧模板的划分为例进行说明。
在一些实施例中，可以只对当前块的上方模板进行划分，对当前块的左侧模板不进行划分，例如将当前块的上方模板划分为M-1个子模板，即P＝M-1，将当前块的左侧模板作为一个子模板，进而得到M个子模板。
在一些实施例中，可以只对当前块的左侧模板进行划分，对当前块的上方模板不进行划分，例如将当前块的左侧模板划分为M-1个子模板，即Q＝M-1，将当前块的上方模板作为一个子模板，进而得到M个子模板。
在一些实施例中，对当前块的上方模板和左侧模板均进行划分，例如将当前块的上方模板划分为P个子模板，将当前块的左侧模板划分为Q个子模板，且M＝P+Q，进而将当前块的模板划分为M个子模板。
在实现方式2中，将当前块的上方模板划分为P个子模板和/或将当前块的左侧模板划分为Q个子模板的方式不做限制，例如可以均等划分，或者按照预设的比例进行划分，或者按照预设的像素点数进行划分，或者按照预设的像素行数或像素列数进行划分等。
在一些实施例中，上述S102-A11中将当前块的左侧模板划分成P个子模板的方式包括但不限于如下几种：
方式1，沿着竖直方向，将上方模板划分为P个子模板。
在一种示例中，沿着竖直方向，将当前块的上方模板平均划分成P等份，例如图18A所示，将当前块的上方模板平均划分成2等份，得到两个子模板，即P＝2。可选的，还可以将当前块的上方模板平均划分为3等份、4等份、5等份等，也就是说，本申请实施例对P的具体取值不做限制，具体根据实际需要确定。需要说明的是，若当前块的上方模板所包括的像素列数不等于P的整数倍时，则可以将P-1个子模板的大小划分为一致，剩余的一个子模板的大小与上述P-1个子模板的大小不一致，例如剩余的一个子模板的大小小于上述P-1个子模板的大小，或者剩余的一个子模板的大小大于上述P-1个子模板的大小。
在另一种示例中，沿着竖直方向，按照预设的子模板比例，将当前块的上方模板划分为P个子模板。例如，按照a1：a2的比例，将当前块的上方模板划分为2个子模板。示例性的，a1：a2＝1:1.5，这样按照1:1.5的比例，将当前块的上方模板划分为两个子模板，分别记为子模板1和子模板2，子模板1与子模板2的大小比例为1:1.5。再例如，按照a1：a2：a3的比例，将当前块的上方模板划分为3个子模板。示例性的，a1：a2：a3＝1:1.5:2，这样按照1:1.5:2的比例，将当前块的上方模板划分为3个子模板，分别记为子模板1、子模板2和子模板3，子模板1、子模板2和子模板3的大小比例为1:1.5:2。
方式2，根据预设的像素点数，将上方模板划分成P个子模板。
该方式2中，将预设的像素点数作为一个最小划分单元，对当前块的上方模板进行划分，划分为P个子模板。本申请对预设的像素点的具体排列方式不做限制，例如，该预设的像素点数排列成一个矩形，将这一矩形块作为上方模板的最小划分单元进行划分。
在一些实施例中，将n列像素作为一个最小划分单元，将上方模板划分成P个子模板，n为正整数。示例性的，假设当前块的上方模板的大小为3*16，即当前块的上方模板包括3行像素行和16列像素列，假设n＝4，则可以将当前块的上方模板中每4列像素列划分为一个单元，进而得到4个单元，根据这4个单元，得到P个子模板。例如，将4个单元中每个单元作为一个子模板，得到4个子模板。也就是说，本申请实施例中，将n列像素作为最小划分单元对当前块的上方模板进行划分时，可以将每个最小划分单元划分的模板作为一个子模板。再例如，将上述4个单元中每两个单元作为一个子模板，进而得到2个子模板。也就是说，本申请实施例中，将n列像素作为最小划分单元对当前块的上方模板进行划分时，可以将多个最小划分单元划分的模板作为一个子模板，例如，将最小划分单元划分的相邻两个或多个区域作为一个子模板。
本申请对上述n的具体取值不做限制，例如为预设值。
可选的，当前块的上方模板的长度与当前块的长度相同，这样上述n可以根据当前块的长度确定，例如当前块的长度为n的正整数倍。示例性的，当前块的长度为16时，则该n可以为2、4、8等数值。
本申请实施例中，若需要对当前块的左侧模板进行划分时，则左侧模板的划分方式与上述当前块的上方模板的划分方式可以相同，也可以不同。
在一些实施例中，上述S102-A11中将当前块的左侧模板划分成Q个子模板的方式包括但不限于如下几种：
方式1，沿着水平方向，将左侧模板划分为Q个子模板。
在一种示例中，沿着水平方向，将当前块的左侧模板平均划分成Q等份，例如图18B所示，将当前块的左侧模板平均划分成2等份，得到两个子模板，即Q＝2。可选的，还可以将当前块的左侧模板平均划分为3等份、4等份、5 等份等，也就是说，本申请实施例对Q的具体取值不做限制，具体根据实际需要确定。需要说明的是，若当前块的左侧模板所包括的像素行数不等于Q的整数倍时，则可以将Q-1个子模板的大小划分为一致，剩余的一个子模板的大小与上述Q-1个子模板的大小不一致，例如剩余的一个子模板的大小小于上述Q-1个子模板的大小，或者剩余的一个子模板的大小大于上述Q-1个子模板的大小。
在另一种示例中，沿着水平方向，按照预设的子模板比例，将当前块的左侧模板划分为Q个子模板。例如，按照b1：b2的比例，将当前块的左侧模板划分为2个子模板。示例性的，b1：b2＝1:1.5，这样按照1:1.5的比例，将当前块的左侧模板划分为两个子模板，分别记为子模板3和子模板4，子模板4与子模板4的大小比例为1:1.5。再例如，按照b1：b2：b3的比例，将当前块的左侧模板划分为3个子模板。示例性的，b1：b2：b3＝1:1.5:2，这样按照1:1.5:2的比例，将当前块的左侧模板划分为3个子模板，分别记为子模板3、子模板4和子模板5，子模板3、子模板4和子模板5的大小比例为1:1.5:2。
方式2，根据预设的像素点数，将左侧模板划分成Q个子模板。
该方式2中，将预设的像素点数作为一个最小划分单元，对当前块的左侧模板进行划分，划分为Q个子模板。本申请对预设的像素点的具体排列方式不做限制，例如，该预设的像素点数排列成一个矩形，将这一矩形块作为左侧模板的最小划分单元进行划分。
在一些实施例中，将m行像素作为一个最小划分单元，将左侧模板划分成Q个子模板，m为正整数。示例性的，假设当前块的左侧模板的大小为16*3，即当前块的左侧模板包括16行像素行和4列像素列，假设m＝4，则可以将当前块的左侧模板中每4行像素行划分为一个单元，进而得到4个单元，根据这4个单元，得到Q个子模板。例如，将4个单元中每个单元作为一个子模板，得到4个子模板。也就是说，本申请实施例中，将m行像素作为最小划分单元对当前块的左侧模板进行划分时，可以将每个最小划分单元划分的模板作为一个子模板。再例如，将上述4个单元中每两个单元作为一个子模板，进而得到2个子模板。也就是说，本申请实施例中，将m行像素作为最小划分单元对当前块的左侧模板进行划分时，可以将多个最小划分单元划分的模板作为一个子模板，例如，将最小划分单元划分的相邻两或多个区域作为一个子模板。
本申请对上述m的具体取值不做限制，例如为预设值。
可选的，当前块的左侧模板的宽度与当前块的宽度相同，这样上述m可以根据当前块的宽度确定，例如当前块的宽度为m的正整数倍。示例性的，当前块的宽度为16时，则该m可以为2、4、8等数值。
根据上述方式，将当前块的模板划分成M个子模板之后，执行上述S102-A2的步骤，即根据权重导出模式，将M个子模板对应到K个模板中。
在本申请实施例中，首先根据上述步骤，将当前块的模板划分成多个子模板，例如将当前块的模板划分为M个子模板，接着，确定这M个子模板中每个子模板属于对应到哪个模板，进而将M个子模板对应到K个模板中，实现对模板的精细准确划分。
上述S102-A2中根据权重导出模式，将M个子模板对应到K个模板中的实现方式包括但不限于如下几种：
方式1，根据权重矩阵的分界线，将M个子模板对应到K个模板中。
具体的，假设K等于2，则将靠近第一预测模式的子模板对应到第一模板中，将靠近第二预测模式的子模板对应到第二模板中。例如图18C所示，当前块的上方模板被划分为4个子模板，分别为子模板1、子模板2、子模板3和子模板4，当前块的左侧模板被划分为2个子模板，分别为子模板5和子模板6。如图18C所示，子模板1和子模板2靠近第一预测模式，因此，将子模板1和子模板2对应到第一模板中，子模板3、子模板4、子模板5和子模板6靠近第二预测模式，因此，将子模板3、子模板4、子模板5和子模板6对应到第二模板中。这样，第一模板包括子模板1和子模板2，在模板匹配时，将子模板1和子模板2作为模板来确定当前块的第一预测模式。对应的，第二模板包括子模板3、子模板4、子模板5和子模板6，在模板匹配时，将子模板3、子模板4、子模板5和子模板6作为模板来确定当前块的第二预测模式，进而实现第一预测模式和第二预测模式的准确确定。
在一些实施例中，若权重的分界线将一个子模板划分成两部分，则可以将该子模板对应到第一模板和第二模板中，此时，第一模板和第二模板具有重叠的部分。例如图18D所示，当前块的上方模板被划分为4个子模板，分别为子模板1、子模板2、子模板3和子模板4，当前块的左侧模板被划分为2个子模板，分别为子模板5和子模板6，且权重的分界线将子模板3划分为两部分，也就是说，权重的分界线与子模板3相交，此时，在一些实施例中，可以如上述图18C所示，将子模板3对应到第二模板中，生成的第一模板与第二模板没有重叠部分。在一些实施例中，如图18D所示，则可以将子模板3分别对应到第一模板和第二模板中，即第一模板包括子模板1、子模板2和子模板3，第二模板包括子模板3、子模板4、子模板5和子模板6，此时第一模板和第二模板具有重叠的部分，这样在后续模板匹配时，子模板3即可以用于确定第一预测模式，也可以用于确定第二预测模式，进而丰富了模板划分非方式。
在一些实施例中，若权重的分界线将一个子模板划分为两个部分，则默认将该子模板对应到第一模板或第二模板。例如图18C所示，权重的分界线将子模板3划分为两部分，则默认将子模板对应到第二模板中。在一种可能的实现方式中，若权重的分界线将一个子模板划分为两部分，则将该子模板对应到较小的一个模板中，例如图18E所示，除去子模板3外，第一模板包括子模板1和子模板2，而第二模板包括子模板4、子模板5和子模板6，第二模板的面积远远大于第一模板的面积，此时为例提高模板匹配的准确性，则将子模板3对应到面积较小的第一模板中，以增加第一模板的面积，进而提高了基于第一模板确定第一预测模式的准确性。
在一些实施例中，权重的分界线将一个子模板划分为两个部分，若该子模板在第一预测模式中的区域大于在第二预测模式中的区域，则将该子模板对应到第一模板中，例如图18C所示，子模板3在第一预测模式中的面积大于在第二预测模式中的面积，因此将该子模板3对应到第一模板中。可选的，若该子模板在第二预测模式中的区域大于在第一预测模式中的区域，则将该子模板对应到第二模板中。
方式2，根据子模板中像素点的权重，将M个子模板对应到K个模板中，具体的，上述S102-A2包括如下步骤：
S102-A21、对于M个子模板中的第j个子模板，根据权重导出模式，确定第j个子模板中的第一点关于第i个预测模式的权重，第i个预测模式为K个预测模式中的任一预测模式。
该方式2通过确定子模板中像素点的权重，确定是将该子模板划分为至哪个模板中，例如子模板中像素点的权重与第一预测模式对应的权重相同或基本相同，则将该子模板对应到第一模板中，若该子模板中像素点的权重与第二预测模式对应的权重相同或基本相同，则将该子模板对应到第二模板中。
由于确定M个子模板中每个子模板对应到那个模板的过程相同，为了便于描述，本申请实施例以M个子模板中的第j个子模板为例进行说明，确定其他子模板对应到哪个模板的过程参照该第j个子模板即可。
在一些实施例中，可以根据第j个子模板中若干像素点的权重来判断是将该第j个子模板对应到哪个模板中。例如，计算第j个子模板中不同位置的多个像素点各自的权重，确定这多个像素点的权重的平均值，根据该权重的平均值确定将该第j个子模板对应到哪个模板中。
在一些实施例中，为例降低计算复杂度，则通过确定第j个子模板中的一个像素点，例如第一点的权重，并根据该第一点的权重确定将第j个子模板对应到哪个模板中。
在一种示例中，上述第一点为第j个子模块中的任意一个点。
在一种示例中，上述第一点为第j个子模板与当前块的交界线上的一个点。例如为交界线上的任意一个点，或者为交界线的中点。
在一些实施例中，可以通过确定第j个子模板中的第一点关于K个预测模式中的任意一个预测模式的权重，来确定第j个子模板对应到哪个模板。
在一些实施例中，可以通过确定第j个子模板中的第一点分别关于K个预测模式权重，来确定第j个子模板对应到哪个模板。
其中确定第j个模板中的第一点关于K个预测模式中每个预测模式的权重的方式相同，本申请实施例以确定第一点关于第i个预测模式的权重为例进行说明。
上述S102-A21中确定第j个子模板中的第一点关于第i个预测模式的权重的方式包括但不限于如下示例：
在一种示例中，将当前块的权重矩阵向第j个子模板进行延伸，以使当前块的权重矩阵至少覆盖第j个子模板中的第一点，进而得到第一点的权重。示例性的，假设第j个子模板为子模板2，第一点为子模板2与当前块的交界线的中点，将图18D所示的权重矩阵向子模板2延伸，以覆盖子模板2中的第一点，进而得到第一点关于第一预测模式的权重，假设最大权重为8，最小权重为0，由于当前块中白色区域的预测值100％来自第一预测模式对应的预测值，当前块中黑色区域的预测值100％来自第二预测模式的预测值，因此，可以确定出第一点关于第一预测模式的权重为8，第一点关于第二预测模式的权重为0。
在另一种示例中，通过如下步骤S102-A211和S102-A212的步骤，确定第j个子模板中的第一点关于第i个预测模式的权重，即上述S102-A21包括如下步骤：
S102-A211、根据权重导出模式确定角度索引和距离索引；
S102-A212、根据角度索引和距离索引，确定第j个子模板中的第一点关于第i个预测模式的权重。
该实现方式中，通过权重导出模式，导出第j个子模板中的第一点关于第i个预测模式的权重，具体是，根据权重导出模式，确定角度索引和距离索引，其中角度索引可以理解为权重导出模式导出的各权重的分界线角度索引。示例性的，可以根据上述表2，确定出权重导出模式对应的角度索引和距离索引，例如权重导出模式为27，则对应的角度索引为12，距离索引为3。接着，根据角度索引和距离索引，确定第j个子模板中的第一点关于第i个预测模式的权重。
在一些实施例中，上述S102-A212包括如下步骤：
S102-A2121、根据角度索引、距离索引和当前块的大小，确定第一点的第一参数；
S102-A2122、根据第一点的第一参数，确定第一点关于第i个预测模式的权重。
本实现方式中，根据角度索引、距离索引、模板的大小和当前块的大小，确定模板中各点的权重，进而将模板中每个点的权重组成的权重矩阵，确定为模板权重。
本申请的第一参数用于确定权重。在一些实施例中，第一参数也称为权重索引。
在一种可能的实现方式中，可以根据如下方式确定出偏移量和第一参数：
下面的例子中模板仅用于Y分量，但是应该理解的是，模板可以应用于任意分量，如Y，Cb，Cr或R，G，B等的任意分量。
记选中的第一点为(x，y)，这个第一点关于第i个预测模式的权重导出过程如下：
这个过程的输入有：当前块的宽度nCbW当前块的高度nCbH，具体如图19所示；GPM的“划分”角度索引变量angleIdx；GPM的距离索引变量distanceIdx；分量索引变量cIdx.因为本例中只以亮度举例，本例中cIdx为0，表示亮度分量。
其中，变量nW,nH,shift1,offset1,displacementX,displacementY,partFlip还有shiftHor按如下方法导出:
nW＝(cIdx＝＝0)？nCbW:nCbW*SubWidthC
nH＝(cIdx＝＝0)？nCbH:nCbH*SubHeightC
shift1＝Max(5,17-BitDepth)其中BitDepth是编解码的比特深度
offset1＝1<<(shift1-1)
displacementX＝angleIdx
displacementY＝(angleIdx+8)％32
partFlip＝(angleIdx>＝13&&angleIdx<＝27)？0:1
shiftHor＝(angleIdx％16＝＝8||(angleIdx％16！＝0&&nH>＝nW))？0:1
其中，变量offsetX and offsetY按如下方法导出:
–如果shiftHor的值为0:
offsetX＝(-nW)>>1
offsetY＝((-nH)>>1)+(angleIdx<16？(distanceIdx*nH)>>3:-((distanceIdx*nH)>>3))
–否则(即shiftHor的值为1):
offsetX＝((-nW)>>1)+(angleIdx<16？(distanceIdx*nW)>>3:-((distanceIdx*nW)>>3)
offsetY＝(-nH)>>1
(x，y)位置相对于第一预测模式的权重wValue按如下方法导出:
其中，变量xL和yL按如下方法导出:
xL＝(cIdx＝＝0)？x:x*SubWidthC
yL＝(cIdx＝＝0)？y:y*SubHeightC
上述disLut按表6确定:
表6距离矩阵disLut的定义
idx 0 2 3 4 5 6 8 10 11 12 13 14
disLut[idx] 8 8 8 4 4 2 0 -2 -4 -4 -8 -8
idx 16 18 19 20 21 22 24 26 27 28 29 30
disLut[idx] -8 -8 -8 -4 -4 -2 0 2 4 4 8 8
其中，第一点的第一参数weightIdx根据如下公式确定：
weightIdx＝(((xL+offsetX)<<1)+1)*disLut[displacementX]+(((yL+offsetY)<<1)+1)*disLut[displacementY]
根据上述方法，确定出第一点的第一参数weightIdx后，根据weightIdx确定出第一点(x，y)关于第i个预测模式的权重。
本申请中，上述S102-A2122中根据第一点的第一参数，确定第一点关于第i个预测模式的权重的方式包括的不限于如下几种：
一种方式，根据第一点的第一参数，确定第一点的第二参数；根据第一点的第二参数，确定第一点关于第i个预测模式的权重。
其中第二参数也用于确定权重。在一些实施例中，上述第二参数也称为第一分量下的权重索引，该第一分量可以为亮度分量、色度分量等。
例如，根据如下公式，确定第一点关于第i个预测模式的权重：
weightIdxL＝partFlip？32+weightIdx:32-weightIdx
wTemplateValue[x][y]＝Clip3(0,8,(weightIdxL+4)>>3)
其中，wTemplateValue[x][y]为第一点(x，y)关于第i个预测模式的权重，weightIdxL为第一点(x，y)的第二参数，也成为第一点在第一分量(例如亮度分量)下的权重索引，partFlip为中间变量，根据角度索引angleIdx确定，例如上述所述：partFlip＝(angleIdx>＝13&&angleIdx<＝27)？0:1，也就是说，partFlip的值为1或0，当partFlip为0时，weightIdxL为32–weightIdx，当partFlip为1时，weightIdxL为32+weightIdx，需要说明的是，这里的32只是一种示例，本申请不局限于此。
另一种方式，根据第一点的第一参数、第一预设值和第二预设值，确定第一点关于第i个预测模式的权重。
为了降低第一点权重的计算复杂度，在一种方式中将第一点关于第i个预测模式的权重限定为第一预设值或第二预设值，也就是说，第一点关于第i个预测模式的权重要么为第一预设值，要么是第二预设值，进而降低第一点关于第i个预测模式的权重计算复杂度。
本申请对第一预设值和第二预设值的具体取值不做限制。
可选的，第一预设值为1。
可选的，第二预设值为0。
在一种示例，可以通过如下公式确定出第一点关于第i个预测模式的权重：
wTemplateValue[x][y]＝(partFlip？weightIdx:-weightIdx)>0？1:0
其中，wTemplateValue[x][y]为第一点(x，y)的权重，上述“1:0”中的1为第一预设值，0为第二预设值。
根据上述方法，确定出第j个子模板中的第一点关于第i个预测模式的权重后，执行如下S102-A22的步骤。
S102-A22、根据第j个子模板中的第一点关于第i个预测模式的权重，将第j个子模板对应到K个模板中。
该方式中，通过确定第j个子模板中的第一点关于第i个预测模式的权重，并根据第一点关于第i个预测模式的权重来确定是将该第j个子模板对应到哪个模板中。
在一种可能的实现方式中，若该第一点关于第i个预测模式的权重与该第i个预测模式的权重相同或基本相同时，则将该第j个子模板对应到第i个模板中。
在另一种可能的实现方式中，若第一点关于第i个预测模式的权重大于第一预设值，则将第j个子模板对应到第i个模板中，第i个模板为K个模板中的一个模板。例如，第j个子模板中的第一点关于第一预测模式的权重大于第一预设值，则将该第j个子模板对应到第一模板中。再例如，第j个子模板中的第一点关于第一预测模式的权重小于或等于第一预设值，则将该第j个子模板对应到第二模板中。
本申请对上述第一预测值的具体取值不做限制。
可选的，上述第一预设值为0。
可选的，上述第一预设值为小于权重中值的任意正数，若权重最大值为8，则权重中值为4。
在一些实施例中，若第一点关于第i个预测模式的权重大于第一预设值，且第一点关于第i+1个预测模式的权重也大于第一预设值，此时，可以将第j个子模板对应到第i个模板中，并将第j个子模板对应到第i+1个模板中，此时第i个模板与第i+1个模板具体重叠部分。以K＝2，第一预测值为0为例，假设第j个子模板为图18D中的子模板3， 第一点为子模板3的下边中点，根据上述方法确定出第一点关于第一预测模式的权重大于0，且第一点关于第二预测模式的权重也大于0，此时，可以将子模板3对应到第一模板和第二模板中。
在一些实施例中，若K为2，i为1，则上述S102-A22包括如下几种示例：
示例1，若第一点关于第一预测模式的权重大于或等于第二预设值，则将第j个子模板对应到第一模板中。
例如，第二预设值为权重中值，若权重的最大值为8时，则权重中值为4，若第j个子模板的第一点关于第一预测模式的权重大于或等于权重中值时，则将第j个子模板对应到第一模板中。以图18D中的子模板2为例，根据上述方法确定出子模板2的第一点关于第一预测模式的权重为8，该8大于第二预测值(例如4)，则可以将子模板2对应到第一模板中。
示例2，若第一点关于第一预测模式的权重小于第二预设值，则将第j个子模板对应到第二模板中。
以图18D中的子模板4为例，根据上述方法确定出子模板4的第一点关于第一预测模式的权重为0，该0小于第二预测值(例如4)，则可以将子模板4对应到第二模板中。
上述结合具体的示例，对情况1中根据权重导出模式，确定K个模板的具体实现方式进行介绍，例如根据权重导出模式对应的权重矩阵的分界线，将当前块的模板划分成K个模板，或者将当前块的模板划分为M个子模板，根据权重导出模式，将M个子模板对应到K个模板中。
本申请实施例，除了使用上述情况1的方法确定出K个模板外，还可以根据如下情况2的方式，确定出K个模板。
情况2，上述S102包括如下步骤：
S102-B1、从预设的不同块大小所对应的第一对应关系中，确定当前块对应的目标第一对应关系，第一对应关系包括不同角度索引或不同的权重导出模式与K个模板之间的对应关系；
S102-B2、从目标第一对应关系中，确定权重导出模式对应的K个模板。
由于当前块可能是正方形，也可能是长方形，可能长度比宽度大也可能宽度比长度大，而且比例也有1：2，1：4等可能。图20A和图20B示出了GPM在32x64块和64x32块的权重矩阵，可以看到不同形状下的划分线与块边界的交点并不一样。因为块形状变了但是划分线的角度并不根据块形状变化而变化。比如索引为52的模式，在32x64的块中与当前块左边界有交点，但是在64x32的块中与当前块左边界没有交点，而对应的交点在下边界。也就是说32x64的块中，模式52的黑色部分与当前块的左侧模板有相邻的部分，而在64x32的块中，模式52的黑色部分与当前块的左侧模板没有相邻的部分。
为了提高模板的选择准确性，本申请实施例根据当前块的长度和宽度设置不同的规则。
例如，对长度等于宽度，长度大于宽度，长度小于宽度这3种情况分别设置不同的第一对应关系，每一个第一对应关系可以为上述表5所示的表格，包括该情况下不同角度索引或不同的权重导出模式与K个模板之间的对应关系。
再例如，按照长宽比，如1：4，1：2，1：1，2：1，4：1等分类，为每一个分类设置一个第一对应关系，该第一对应关系包括该分类下不同角度索引或不同的权重导出模式与K个模板之间的对应关系。
这样解码时，解码端可以根据当前块的大小，例如当前块的长度和宽度，从预设的不同块大小所对应的第一对应关系中，确定当前块对应的目标第一对应关系，并根据权重导出模式，从该目标第一对应关系中，获得该权重导出模式对应的K个模板。在一些实施例中，若上述目标第一对应关系包括不同角度索引与K个模板之间的对应关系，则需要根据权重导出模式确定出目标角度索引，再根据目标角度索引从目标第一对应关系中查询该目标角度索引所对应的K个模板。
本申请实施例中，解码端根据上述步骤，确定出K个模板后，执行如下S103的步骤，根据K个模板确定当前块的K个预测模式。
S103、根据K个模板，确定K个预测模式。
本申请实施例中，K个模板中的每个模板用于确定一个预测模式，例如使用K个模板中的第一模板确定第一预测模式，使用K个模板中的第二模板确定第二预测模式。
本申请实施例中使用K个模板中的每个模板确定对应的预测模式的过程一致，本申请实施例以使用K个模板中的第i个模板确定第i个预测模式为例进行说明。
在一些实施例中，上述S103包括如下S103-A1至S103-A4的步骤：
S103-A1、针对K个预测模式中的第i个预测模式，获取至少一个候选预测模式。
上述至少一个候选预测模式可以理解为第i个预测模式对应的候选预测模式，在一些实施例中，不同的预测模式对应的候选预测模式可以不同。在一些实施例中，若两个预测模式的类型相同时，例如均为帧内预测模式时，这两个预测模式对应的候选预测模式可以相同。
本申请实施例中，解码端在确定第i个预测模式时，首先判断该第i个预测模式是否是通过模板匹配方式确定的。
在一种可能的实现方式中，码流中携带一标志A，该标志A用于指示第i个预测模式是否通过模板匹配方式确定。示例性的，若该标志A的取值为1时，说明第i个预测模式是通过模板匹配的方式确定，若该标志的取值为0时，说明该第i个预测模式不是通过模板匹配的方式确定。
基于此，解码端解码码流，得到该标志A，并判断该标志A的取值，若该标志A的取值为1时，则确定第i个预测模式为通过模板匹配方式确定的，此时，则解码端执行本申请实施例的方法，获取至少一个候选预测模式，并确定候选预测模式的代价，根据候选预测模式的代价，确定第j个预测模式。
在另一种可能的实现方式中，编码端和解码端均默认该第j个预测模式是通过模板匹配方式确定的，这样解码端在确定第j个预测模式时，默认使用模板匹配方式确定第j个预测模式，接着获取至少一个候选预测模式，并确定候选预测模式的代价，根据候选预测模式的代价，确定第j个预测模式。
在一些实施例中，若上述第j个预测模式为帧间预测模式时，则上述至少一个候选预测模式包括一个或多个帧间 预测模式，例如包括skip、merge、普通帧间预测模式、单向预测、双向预测、多假设预测等中的至少一个。
在一些实施例中，若上述第j个预测模式为帧内预测模式时，则上述至少一个候选预测模式包括直流(Direct Current，DC)模式、平面(PLANAR)模式、角度模式等中的至少一个。可选的，上述至少一个候选预测模式包括MPM列表中的帧内预测模式。
在一些实施例中，至少一个候选预测模式还可以包括IBC、palette等模式。
本申请对至少一个候选预测模式所包括的预测模式的类型以及预测模式的个数不做限制。
可选的，上述至少一个候选预测模式为预设模式。
可选的，上述至少一个候选预测模式为MPM列表中的模式。
可选的，上述至少一个候选预测模式是根据一些规则，如等间距筛选等，确定出的候选预测模式的集合。
S103-A2、使用候选预测模式对第i个模板进行预测，得到第i个模板的预测值；
示例性的，针对至少一个候选预测模式中的每一个候选预测模式，使用该候选预测模式对第i个模板进行预测，确定第i个模板的预测值。
其中，该第i个模板的预测值可以理解为由第i个模板中每个像素点的预测值组成的矩阵。
S103-A3、根据第i个模板的预测值和重建值，确定候选预测模式的代价。
示例性的，针对至少一个候选预测模式中的每一个候选预测模式，根据每个候选预测模式关于第i个模板的预测值，以及第i个模板的重建值，确定每个候选预测模式的代价。例如，根据候选预测模式关于第i个模板的预测值和第i个模板的重建值确定该候选预测模式对于第i个模板的损失，根据该候选预测模式对于第i个模板的损失，确定出候选预测模式的代价。
上述S103-A3中确定候选预测模式的代价的方式包括但不限于如下几种：
方式一，采用矩阵的方式确定候选预测模式的代价。具体是，根据候选预测模式关于第i个模板的预测值和第i个模板的重建值，确定损失样本，由于上述候选预测模式关于第i个模板的预测值和第i个模板的重建值均为矩阵，因此得到损失样本也为矩阵。例如将候选预测模式关于第i个模板的预测值与该第i个模板的重建值的差值的绝对值确定为损失样本。接着，根据该损失样本，确定候选预测模式关于第i个模板的代价，例如将该损失样本中每个点的损失之和确定为该候选预测模式关于第i个模板的代价。
方式二，采用逐点计算的方式，确定候选预测模式的代价，即上述S103-A3包括如下步骤：
S103-A321、针对第i个模板中的第i个点，确定第i个点在第i个模板的预测值中对应的第i个预测值，与在第i样本的重建值中对应的第i个重建值之间的损失；
S103-A322、根据第i个点对应的损失，确定候选预测模式在第i个点处的代价；
S103-A323、根据候选预测模式在第i个模板中各点处的代价，确定候选预测模式的代价。
上述第i个点可以理解为第i个模板中的任意一个点，也就是说，确定第i个模板中每个点的代价的过程相同，参照第i个点即可。具体是，使用候选预测模式对第i个模板进行预测，得到候选预测模式关于第i个模板的预测值，将第i个点在第i个模板的预测值中对应的预测值记为第i个预测值，将该第i个点在第i个模板的重建值中对应的重建值记为第i个重建值，进而根据第i个预测值和第i个重建值，确定出该候选预测模式在第i个点的损失，并根据该候选预测模式在第i个点的损失，确定候选预测模式在第i个点处的代价，例如将该候选预测模式在第i个点的损失确定为该候选预测模式在第i个点处的代价。根据上述方法，确定出该候选预测模式在第i个模板中的每个点或多个点处的代价，进而根据第i个模板中每个点或多个点的代价，确定出候选预测模式关于第i个模板的代价。例如，将候选预测模式在第i个模板中各点处的代价之和，确定为候选预测模式关于第i个模板的代价，或者将候选预测模式在第i个模板中各点处的代价平均值，确定为候选预测模式关于第i个模板的代价，本申请对根据第i个模板中至少一个点的代价，确定候选预测模式关于第i个模板的代价不做限制。
示例性的，以SAD代价为例，可以根据如下公式(1)确定出候选预测模式在第i个模板中的第i个点(x，y)处的代价：
tempValueA[x][y]＝abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y])   (1)
示例性的，根据如下公式(2)确定出候选预测模式的代价：
costCandA＝∑tempValueA[x][y]     (2)
其中，abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y])是第i个模板中点(x，y)的预测值predTemplateSamplesCandA和重建值recTemplateSamples的差的绝对值，将该差的绝对值称为点(x，y)对应的损失。tempValueA[x][y]可以认为是该候选预测模式在这个点(x，y)的代价。该候选预测模式在第i个模板上的总的代价costCandA为第i个模板上每一个点的代价累加。
需要说明的是，上述以SAD为例来确定候选预测模式的代价，可选的还可以根据SATD、MSE等代价计算方法，确定候选预测模式关于第i个模板的代价。
根据上述方法，可以确定出候选预测模式关于第i个模板的代价，接着执行如下S103-A4的步骤。
S103-A4、根据至少一个候选预测模式的代价，确定第i个预测模式。
本申请实施例，若第i个预测模式为通过模板匹配方法确定，则通过上述方法，确定出候选预测模式的代价，并根据候选预测模式的代价，确定第i个预测模式。
示例1，将至少一个候选预测模式中代价最小的候选预测模式，确定为第i个预测模式。
示例2，根据候选预测模式的代价，从至少一个候选预测模式中选出一个或多个候选预测模式；根据一个或多个候选预测模式，确定第j个预测模式。
在该示例2的一种可能的实现方式中，解码端从一个或多个候选预测模式中选出一个候选预测模式，作为第j个 预测模式。
具体是，根据编码端的指示，从上述一个或多个候选预测模式中，确定出第i个预测模式。例如，上述一个或多个候选预测模式为M个，编码端将这M个候选预测模式按照代价进行排序，例如按照代价从小到大对M个候选预测模式进行排序，或者按照代价从大到小对M个候选预测模式进行排序，从排序后的M个候选预测模式中确定出一个候选预测模式B作为第i个预测模式。同时，编码端将该候选预测模式模式B的标识编入码流中，该候选预测模式B的标识可以是候选预测模式B在M个候选预测模式中的排序号，也可以是该候选预测模式B的模式索引号。这样，解码端通过解码码流，得到该候选预测模式B的标识，进而根据候选预测模式B的标识，将上述确定的M个候选预测模式中候选预测模式B的标识对应的候选预测模式，确定为第i个预测模式。
在该示例2的另一种可能的实现方式中，解码端获取当前块的备选预测模式；确定备选预测模式对第i个模板进行预测时的代价；根据备选预测模式对第i个模板进行预测时的代价以及上述选出的一个或多个候选预测模式关于第i个模板的代价，从备选预测模式和上述一个或多个候选预测模式中选出一个预测模式，作为第i个预测模式。
可选的，上述当前块的备选预测模式包括当前块周围已重建解码块的预测模式和/或预设预测模式中的一个或多个。
可以理解的是，在本申请中，预设预测模式可以包括DC模式、Bilinear模式、Planar模式等多种不同模式中的一种或多种。
具体的，解码端获取当前块的备选预测模式，例如将当前块周围已重建解码块的预测模式和/或预设预测模式中的一个或多个作为当前块的备选预测模式。接着，确定每个备选预测模式对模板进行预测时的代价，例如使用备选预测模式对当前块进行预测，得到预测值，将该预测值与模板的重建值进行比较，得到该备选预测模式的代价，其中备选预测模式的代价可以是SAD、SATD等代价。根据备选预测模式的代价以及上述一个或多个候选预测模式的代价，从备选预测模式和上述一个或多个候选预测模式中选出一个预测模式作为第j个预测模式，例如将备选预测模式和上述一个或多个候选预测模式中代价最小的预测模式，确定为第j个预测模式。
值得注意的是，上述当前块的备选预测模式与上述确定的一个或多个候选预测模式不相同，也就是说，解码端将当前块周围已重建解码块的预测模式和/或预设预测模式中与上述一个或多个候选预测模式中相同的预测模式删除，将剩余的预测模式确定为当前块的备选预测模式。
可以理解的是，对帧间预测来说，模板匹配可以在一个初始运动信息的基础上进行“搜索”。一个预测模式需要确定一个运动信息。可以在一个初始运动信息的周边一定范围内确定一些运动信息，从而确定一些预测模式。如给定一个初始运动信息，其运动矢量是(xInit，yInit)，设置一个搜索范围如水平方向从xInit-sR到xInit+sR，竖直方向从yInit-sR到yInit+sR的矩形区域，其中sR可以是2，4，8等。该矩形区域内的每一个运动矢量都可以和初始运动信息的其他信息，如参考帧索引和预测列表标志等组合起来确定一个运动信息，从而确定一个预测模式。上述至少一个候选预测模式可以包含所述确定的预测模式。比如如果GPM在merge模式下使用，如果使用模板匹配的方法确定第一预测模式，可以使用merge_gpm_idx0从mergeCandList中确定一个初始运动信息。再根据上述方法确定(2*sR+1)*(2*sR+1)个运动信息，从而确定一些预测模式，这些预测模式都是merge模式，或者称为使用模板匹配的merge模式。
值得注意的是，如果上述至少一个候选预测模式组成的集合包含的预测模式很多，出于复杂度的考虑，可能不会对至少一个候选预测模式中的每一个候选预测模式都确定其代价。在一些实施例中，根据至少一个候选预测模式的代价，确定第j个预测模式的过程也可以进一步扩展为几层的由粗选到细选的过程。比如帧间预测模式中，运动矢量支持分像素精度，如1/4、1/8、1/16精度等。所以可以先从包含整像素运动矢量的预测模式中选出代价最小的预测模式，再在该预测模式和运动矢量在该模式的运动矢量附近的包含分像素运动矢量的预测模式中进一步选出代价最小的预测模式。比如在帧内预测模式中，根据候选预测模式的代价，先按一定粒度选出一个或几个帧内预测模式，再在该一个或几个帧内预测模式及更细粒度的相邻帧内预测模式中再进行筛选。
本申请实施例中，若K个预测模式中的第i个预测模式通过模板匹配方式确定时，通过获取至少一个候选预测模式，并使用候选预测模式对模板进行预测，得到该候选预测模式下模板的预测值；根据该候选预测模式下模板的预测值和模板的重建值，得到该候选预测模式的代价，最后根据候选预测模式的代价，得到第j个预测模式。
上述实施例以K个预测模式中的第i个预测模式的确定过程为例进行说明，K个预测模式的中的其他预测模式的确定过程与上述第i个预测模式的确定过程一致，参照即可。例如，K＝2，根据上述方法确定出第一预测模式和第二预测模式，接着，解码端根据该第一预测模式得到第一预测值，根据第二预测模式得到第二预测值，将第一预测值和第二预测值进行加权，得到新的预测值。
本申请实施例根据上述方法，可以根据K个模板确定出K个预测模式，接着使用这K个预测模式对当前块进行预测，得到当前块的预测值，具体参照下面的S104所述。
S104、根据K个预测模式和权重导出模式，确定预测值。
本申请，根据权重导出模式确定权重，根据K个预测模式确定K个预测值，根据权重对K个预测值进行加权，将加权结果确定为最终的预测值。
在本申请中，权重导出模式用于确定当前块的预测值进行加权时的权重。具体地，权重导出模式可以是导出权重的模式。对于一个给定长度和宽度的块，每一种权重导出模式可以导出一个权重矩阵；对于同样大小的块，不同权重导出模式导出的权重矩阵可以不同。
示例性的，在本申请中，AVS3的AWP有56种权重导出模式，VVC的GPM有64种权重导出模式。
可以理解的是，在本申请的实施例中，解码端在基于K个预测模式以及权重，确定预测值时，可以先根据K个预测模式中的每个预测模式，确定每个预测模式对应的预测值，对每个预测模式对应的预测值进行加权，得到最终的预测值。例如K＝2，使用第一预测模式确定第一预测值，使用第二预测模式确定第二预测值，然后可以利用权重对第一预测值和第二预测值进行加权计算，便可以获得最终的预测值。
在一些实施例中，上述预测过程是以像素点为单位进行的，对应的上述权重也为像素点对应的权重。此时，对当前块进行预测时，使用K个预测模式中的每个预测模式对当前块中的某一个像素点A进行预测，得到K个预测模式关于像素点A的K个预测值，根据像素点A的权重对这K个预测值进行加权，得到像素点A的最终预测值。对当前块中的每一个像素点执行上述步骤，可以得到当前块中每个像素点的最终预测值，当前块中每个像素点的最终预测值构成当前块的最终预测值。以K＝2为例，使用第一预测模式对当前块中的某一个像素点A进行预测，得到该像素点A的第一预测值，使用第二预测模式对该像素点A进行预测，得到该像素点A的第二预测值，根据像素点A对应的权重，对第一预测值和第二预测值进行加权，得到像素点A的最终预测值。
在一种示例中，以K＝2为例，若第一预测模式和第二预测模式均为帧内预测模式时，采用第一帧内预测模式进行预测，得到第一预测值，采用第二帧内预测模式进行预测，得到第二预测值，根据预测权重对第一预测值和第二预测值进行加权，得到新的预测值。例如，采用第一帧内预测模式对像素点A进行预测，得到像素点A的第一预测值，采用第二帧内预测模式对像素点A进行预测，得到像素点A的第二预测值，根据像素点A对应的预测权重，对第一预测值和第二预测值进行加权，得到像素点A的最终预测值。
在一些实施例中，若K个预测模式中的第i个预测模式为帧间预测模式，则上述根据K个预测模式和权重导出模式，确定预测值包括如下步骤：
S104-AB21、根据第i个预测模式，确定运动信息；
S104-AB22、根据运动信息，确定第i个预测值；
S104-AB23、根据K个预测模式中除第i个预测模式外的其他预测模式，确定K-1个预测值；
S104-AB24、根据权重导出模式，确定权重；
S104-AB25、根据第i个预测值、K-1个预测值和权重，确定预测值。
以K＝2为例，若第一预测模式为帧内预测模式，第二预测模式为帧间预测模式时，采用帧内预测模式进行预测，得到第一预测值，采用帧间预测模式进行预测，得到第二预测值，根据预测权重对第一预测值和第二预测值进行加权，得到新的预测值。在该示例中，采用帧内预测模式对当前块中每一个点进行预测，得到当前块中每一个点的预测值，当前块中每一个点的预测值，构成当前块的第一预测值。采用帧间预测模式，确定一运动信息，根据该运动信息确定当前块的最佳匹配块，将该最佳匹配块确定为当前块的第二预测值。针对当前块中每个像素点的预测权重，对当前块的第一预测值和第二预测值进行逐点加权运算，得到当前块的新的预测值。例如，对于当前块中的像素点A，根据像素点A的预测权重，将当前块的第一预测值中像素点A对应的第一预测值，与当前块的第二预测值中像素点A对应的第二预测值进行加权，得到像素点A的最终预测值。
在一些实施例中，若K大于2时，则可以根据权重导出模式确定K个预测模式中两个预测模式的预测权重，K个预测模式中的其他预测模式的预测权重可以为预设值。例如，K＝3，第一预测模式和第二预测模式的预测权重根据权重导出模式导出，第三预测模式的预测权重为预设值。在一些实施例中，若K个预测模式的总预测权重一定，例如为8，则可以根据预设权重比例，来确定K个预测模式各自的预测权重，假设第三预测模式的预测权重占整个预测权重的1/4，则可以确定第三预测模式的预测权重为2，剩下3/4的预测权重分配给第一预测模式和第二预测模式。示例性的，如果根据权重导出模式导出第一预测模式的预测权重3，则确定第一预测模式的预测权重为(3/4)*3，第二预测模式的预测权重为第一预测模式的预测权重为(3/4)*5。
在一些实施例中，在执行本申请实施例的方法之前，解码端需要判断当前块是否适用于模板匹配方法，若解码端确定当前块适用于模板匹配方法，则执行上述S101至S104的步骤，若解码端确定当前块不适用于模板匹配方法时，则使用其他的方式确定K个预测模式。
示例性的，解码端通过如下几种方式，确定当前块是否适用于模板匹配方法：
方式一，解码端根据K个模板所包括的点数，确定当前块是否适用于模板匹配方法。
由于目前的编解码顺序是从左到右，从上到下的，这导致当前块能获得的模板是在当前块的左侧和上测，而右侧和下侧是不可得的，而诸如右上方和左下方，有的情况是可得的有的情况是不可得的。可以注意到GPM的某些权重导出模式下，一个预测模式找不到对应的模板或已重建的相邻区域。比如说正方形块的情况下的GPM的索引为55，56，57的权重矩阵，其白色区域只存在于右下角，没有与白色区域直接相邻的模板或相邻的已重建区域。另外还有一些模式，虽然可以找到直接相邻的模板或相邻的已重建区域，但是相邻的区域非常小，如正方形块的情况下的GPM的索引为59，60的权重矩阵中的白色区域。本申请实施例将与当前块直接相邻的模板或直接相邻的重建区域称作可用区域，如果没有找到可用区域或者可用区域非常小，而强行将对应的预测模式应用模板匹配或纹理特性方法不仅不会提高压缩效率反而可能起到反作用。因为这个预测模式跟整个或大部分模板或相邻重建区域的特性是不同的。
也就是说，本申请实施例中，可用模板比较大的预测模式使用模板匹配或相邻重建像素的纹理特性，对可用模板比较小的预测模式则不使用模板匹配或相邻重建像素的纹理特性。例如图17B中，第一预测模式对应的可用模板为模板中的白色区域和灰色区域，第二预测模式对应的可用模板为模板中的黑色区域和灰色区域，由图17B可知，第一预测模式对应的可用模板的面积较大，例如大于预设值，则解码端确定第一预测模式适用于模板匹配方法，同理，由图17B可知，第二预测模式对应的可用模板的面积较大，例如大于预设值，则解码端确定第二预测模式也适用于模板匹配方法。
在一种可能的实现方式中，可以根据K个模板所包括的点数，确定当前块是否适用于模板匹配方法。
在一种示例中，若K个模板所包括的点数均大于预设阈值时，确定当前块适用于模板匹配方法，进而执行上述S103根据K个模板，确定K个预测模式。
可选的，上述预设阈值可以是0。
可选的，上述预设阈值为权重中值，例如4。
可选的，上述预设阈值为一个定值。
可选的，上述预设阈值根据当前块的尺寸确定，例如为当前块的总点数的1/m1，m1为正数。
可选的，上述预设阈值根据当前块的模板的尺寸确定，例如为当前块的模板的总点数的1/m2，m2为正数。
在另一种示例中，若K个模板中至少一个模板包括的点数小于预设阈值时，则根据权重导出模式，确定K个预测模式。
本申请实施例中，解码端根据上述S102的步骤，基于当前块的大小和所述权重导出模式中的至少一个，确定K个模板后，根据K个模板所包括的点数，确定当前块是否适用于模板匹配方法。具体是，针对K个模板中的第i个模板，若该第i个模板所包括的像素点的个数大于预设阈值时，则说明第i个模板中用于确定第i个预测模式的可用模板较大，使用该第i个模板确定第i个预测模式时，可以提高预测效果。若该第i个模板所包括的像素点的个数小于预设阈值时，则说明该第i个模板中用于确定第i个预测模式的可用模板较小或不存在，此时使用模板匹配方法确定第i个预测模式时，不仅不会提高压缩效率反而可能起到反作用。
方式二，解码端码流，得到第一标志，该第一标志用于指示是否采用模板匹配方式导出预测模式；进而根据该第一标志确定当前块是否采用模板匹配方式导出预测模式。
在该方式二中，若编码端将第一标志写入码流，该第一标志用于指示当前块是否采用模板匹配方式导出预测模式，若编码端确定当前块采用模板匹配方式导出预测模式时，则将该第一标志置为1，且将置为1的第一标志写入码流，若编码端确定当前块不采用模板匹配方式导出预测模式时，则将该第一标志置为0，且将置为0的第一标志写入码流。这样，解码端获得码流后，通过解码码流，得到该第一标志，并且根据该第一标志确定当前块是否采用模板匹配方式导出预测模式。
示例性的，若第一标志指示采用模板匹配方式导出预测模式时，则根据权重导出模式，确定K个预测模式。例如，解码端解码码流，得到第一标志，当第一标志的取值为1时，解码端确定当前块采用模板匹配方式导出预测模式，进而执行上述S102的步骤，根据权重导出模式，确定K个预测模式。
示例性的，若第一标志指示不采用模板匹配方式导出预测模式时，则通过权重导出模式，确定当前块的K个预测模式中的至少一个。例如，解码端解码码流，得到第一标志，当第一标志的取值为0时，解码端确定当前块不采用模板匹配方式导出预测模式，进而根据其他方式确定K个预测模式，例如通过权重导出模式，确定当前块的K个预测模式中的至少一个。
在本申请中，权重变化的位置构成一条直线(曲线段)，或，如图4和图5所示过渡区域中权重相同的位置构成一条直线(曲线段)。可以将这条直线叫做分界线(或划分线或分割线)。分界线本身也是有角度的，可以设水平向右的角度为0，角度逆时针增加。那么分界线可能有水平0度，竖直90度，倾斜的如45度，135度，以及其他各种不同的角度等。如果一个块选择使用某一个权重矩阵，那么对应的纹理很可能在分界线两边显现出不同的特性，比如分界线两边是两种不同角度的纹理，或者，分界线的一边是一种角度的纹理，而另一边是一种比较平坦的纹理。由于分界线本身也是有角度的，因此可以假设一个点经过角度预测得到的，它可能与当前块的某些纹理是接近的，因而这条直线与当前块的两个预测模式是存在相关性的。
具体地，在本申请中，假设分界线是由一个点经过角度预测得到的，那么可以找到至少一个角度预测模式，这个角度预测模式可以近似地做出分界线。例如，水平方向的分界线就匹配水平预测模式，如在VVC中的模式18；竖直方向的分界线就匹配竖直帧内预测模式，如在VVC中的模式50。45度的分界线可以匹配左下到右上45度的帧内预测模式，如VVC中的模式66，也可以匹配右上到左下225度的帧内预测模式，如VVC中的模式2。那么权重导出模式就可以匹配到某些帧内预测模式。
需要说明的是，在本申请中，权重导出模式也可以为权重的索引，如AWP的56种模式就可以认为是56种权重导出模式，VVC的GPM的64种模式就可以认为是64种权重导出模式。
在一些实施例中，除了与权重分界线对应的帧内角度预测模式使用的可能性高之外，和它相关的某些帧内角度预测模式使用的可能性也较高。比如说与这个分界线相近的角度，或这个分界线垂直的角度等对应的帧内角度预测模式。
在一些实施例中，如果K个预测值都由帧内预测模式预测得到，那么GPM要使用K个不同的帧内预测模式。
在一些实施例中，如果K个预测值其中至少一个由帧内预测模式预测而来，至少一个由其他预测方法而来，那么GPM需要使用一个或少数几个帧内预测模式，这种情况下就可以提供一个更小的范围的帧内预测模式供GPM选择，以达到节约选择哪一个帧内预测模式的标志的开销的目的。
在一些实施例中，如果GPM的一个预测值来自于帧内预测，一个预测值来自于帧间预测。若本申请所使用的帧内预测模式默认由权重导出模式确定。比如权重导出模式的分界线是水平方向的，如图4中GPM的索引为18，19，50，51的模式，就确定帧内预测模式为水平方向的模式18。比如权重导出模式的分界线是竖直方向的，如图4中GPM的索引为0，1，36，37的模式，就确定帧内预测模式为垂直方向的模式50。
也就是说，本申请在根据权重导出模式确定K个预测模式中的至少一个之前，首先要确定K预测模式的类型，当预测模式为帧内预测模式时，才可以根据权重导出模式确定该预测模式。
在此基础上，在根据权重导出模式确定K预测模式中的至少一个之前，本申请实施例的方法还包括：
步骤11-0、解码码流，得到类型标志，该类型标志用于指示K预测模式分别是否属于帧内预测模式；
步骤11-1、根据类型标志，确定K个预测模式的类型。
下面以K＝2为例。
示例性的，若类型标志的取值为第一数值时，指示第一预测模式和第二预测模式均为帧间预测模式，此时，mode0IsInter为1，mode1IsInter为1，其中mode0IsInter指示第一预测模式是否为帧间预测模式，mode1IsInter指示第二预测模式是否为帧间预测模式，当第一预测模式为帧间预测模式时，mode0IsInter为1，当第二预测模式为帧间预测模式时，mode1IsInter为1。
示例性的，类型标志的取值为第二数值时，指示第一预测模式为帧内预测模式，第二预测模式为帧间预测模式，此时，mode0IsInter为0，mode1IsInter为1。
示例性的，类型标志的取值为第三数值时，指示第一预测模式为帧间预测模式，第二预测模式为帧内预测模式，此时，mode0IsInter为1，mode1IsInter为0。
示例性的，类型标志的取值为第四数值时，指示第一预测模式和第二预测模式均为帧内预测模式，此时，mode0IsInter为0，mode1IsInter为0。
本申请对上述第一数值、第二数值、第三数值和第四数值的具体取值不做限制。
可选的，第一数值为0。
可选的，第二数值为1。
可选的，第三数值为2。
可选的，第四数值为3。
在一种示例中，可以用字段intra_mode_idx表示类型标志。
本申请中，编码端根据类型标志确定出第一预测模式和第二预测模式的类型后，在编码时，需要将该类型标志编入码流，解码端解码码流，得到该类型标志，并根据类型标志确定第一预测模式和第二预测模式的类型。
可选的，如上述表4的方式，其中，merge_gpm_partition_idx为权重导出模式或权重导出索引，intra_mode_idx为类型标志，merge_gpm_idx0是第一个运动信息在候选列表中的索引值，merge_gpm_idx1是第二个运动信息在候选列表中的索引值。
本申请中，解码端根据上述类型标志，确定出K预测模式的类型后，若K预测模式中至少一个为帧内预测模式时，基于权重导出模式确定该帧内预测模式。
也就是说，本申请中基于权重导出模式来确定帧内预测模式，例如，第一预测模式和第二预测模式均为帧内预测模式时，则基于权重导出模式来确定第一预测模式和第二预测模式。再例如，第一预测模式和第二预测模式中的一个为帧内预测模式时，则基于权重导出模式确定第一预测模式和第二预测模式中的帧内预测模式。
本申请中，基于权重导出模式确定K个预测模式中的至少一个的方式包括但不限于如下几种：
方式一，若K个预测模式中的至少一个为帧内预测模式时，则根据权重导出模式确定角度索引，并将角度索引对应的帧内预测模式，确定为K个预测模式中的一个预测模式。
其中角度索引用于指示权重的分界线角度索引。
在一些实施例中，用字段angleIdx表示角度索引。
上述表2示出了merge_gpm_partition_idx与angleIdx的对应关系，参照上述表2，可以根据该权重导出模式导出角度索引。
本申请中，角度索引与帧内预测模式具有对应关系，即不同的角度索引与不同帧内预测模式对应。
示例性的，角度索引与帧内预测模式具有对应关系如表7所示：
表7
angleIdx 帧内预测模式
0 50
2 42
3 38
4 34
5 30
… …
该方式一中，以K＝2为例，若第一预测模式或第二预测模式为帧内预测模式时，则根据权重导出模式确定角度索引，例如根据上述表2，导出权重导出模式对应的角度索引。接着，在上述表7中，确定出该角度索引对应的帧内预测模式，例如，角度索引为2，其对应的帧内预测模式为42，进而将帧内预测模式42确定为第一预测模式或第二预测模式。
方式二，若K个预测模式中的至少一个为帧内预测模式时，则获取权重导出模式对应的帧内预测模式；根据权重导出模式对应的帧内预测模式，确定K个预测模式中的至少一个。
该方式二中，以K＝2为例，若第一预测模式和/或第二预测模式为帧内预测模式时，第一预测模式和/或第二预测模式是从权重导出模式对应的帧内预测模式中确定出。例如，第一预测模式和/或第二预测模式可以是跟权重划分线(也称为分界线)同一条直线的或近似同一条直线的帧内预测模式。或者，第一预测模式和/或第二预测模式可以是跟权重划分线垂直的或近似垂直的帧内预测模式。例如，权重的分界线是水平方向的，如图4中GPM的索引为18，19，50，51的模式，第一预测模式和/或第二预测模式为水平方向的模式18和垂直方向的模式50。
由上述可知，与权重导出模式对应的帧内预测模式的类型较多，例如包括与权重的分界线平行的帧内预测模式、与分界线垂直的帧内预测模式等。本申请可以通过标志来指示第一预测模式和/或第二预测模式具体选择权重导出模式对应的帧内预测模式中的那一个模式。
示例性的，以K＝2为例，若第一预测模式为帧内预测模式，则使用第二标志来指示第一预测模式与权重导出模式对应的帧内预测模式的对应关系，例如第二标志指示第一预测模式为与权重的分界线平行的帧内预测模式，或者指示第一预测模式为与权重的分界线垂直的帧内预测模式。
示例性的，若第二预测模式为帧内预测模式，则使用第三标志来指示第二预测模式与权重导出模式对应的帧内预测模式的对应关系，例如第三标志指示第二预测模式为与权重的分界线平行的帧内预测模式，或者指示第二预测模式为与权重的分界线垂直的帧内预测模式。
基于此，上述方式二中根据权重导出模式对应的帧内预测模式，确定第一预测模式和/或第二预测模式的方式包括但不限于如下几种示例：
示例1，若第一预测模式为帧内预测模式时，则获取第二标志，并将权重导出模式对应的帧内预测模式中第二标 志对应的帧内预测模式，确定为第一预测模式。
示例2，若第二预测模式为帧内预测模式时，则获取第三标志，并将权重导出模式对应的帧内预测模式中第三标志对应的帧内预测模式，确定为第二预测模式。
在一些实施例中，权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式和与分界线垂直的帧内预测模式中的至少一个。
可选的，当第二标志为第五数值，例如0时，指示第一预测模式为权重导出模式对应的帧内预测模式中与权重的分界线平行的帧内预测模式。当第二标志为第六数值，例如1时，指示第一预测模式为权重导出模式对应的帧内预测模式中与权重的分界线垂直的帧内预测模式。
可选的，当第三标志为第五数值，例如0时，指示第二预测模式为权重导出模式对应的帧内预测模式中与权重的分界线平行的帧内预测模式。当第三标志为第六数值，例如1时，指示第二预测模式为权重导出模式对应的帧内预测模式中与权重的分界线垂直的帧内预测模式。
在一些实施例中，权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式、与分界线垂直的帧内预测模式和planar模式中的至少一个。
可选的，当第二标志为第五数值，例如0时，指示第一预测模式为权重导出模式对应的帧内预测模式中与权重的分界线平行的帧内预测模式。当第二标志为第六数值，例如1时，指示第一预测模式为权重导出模式对应的帧内预测模式中与权重的分界线垂直的帧内预测模式。当第二标志为第七数值，例如2时，指示第一预测模式为planar模式。
可选的，当第三标志为第五数值，例如0时，指示第二预测模式为权重导出模式对应的帧内预测模式中与权重的分界线平行的帧内预测模式。当第三标志为第六数值，例如1时，指示第二预测模式为权重导出模式对应的帧内预测模式中与权重的分界线垂直的帧内预测模式。当第三标志为第七数值，例如2时，指示第二预测模式为planar模式。
在一种示例的，用字段intra_gpm_idx0表示第二标志。
在一种示例中，用字段intra_gpm_idx1表示第三标志。
本申请中，若第一预测模式为帧内预测模式时，则根据上述第二标志确定出第一预测模式，若第二预测模式为帧内预测模式时，则根据上述第三标志，确定第二预测模式。
示例性的，第二标志(intra_gpm_idx0)和/或第三标志(intra_gpm_idx1)如表8所示。
表8
Figure PCTCN2021143977-appb-000005
解码端解码表8所示的码流，得到第二标志和/或第三标志，并根据该第二标志和/或第三标志，确定第一预测模式和/或第二预测模式，进而使用第一预测模式和第二预测模式以及权重，确定预测值。
在一些实施例中，若第一预测模式和第二预测模式均为帧内预测模式时，则第二标志与第三标志的取值不相同。为了使得第二标志和第三标志的取值不同，一种可行的方式是，使第二标志(intra_gpm_idx1)的取值为0和1，如果intra_gpm_idx1大于intra_gpm_idx0，令intra_gpm_idx1加1。
参照上述方法，根据权重导出模式确定K个预测模式中的至少一个，根据K个预测模式确定K个预设值，对这K个预测值进行加权处理，得到最终的预测值。
本申请实施例提供的预测方法，解码端通过解码码流，确定当前块的权重导出模式；根据当前块的大小和权重导出模式中的至少一个，确定K个模板；根据K个模板，确定K个预测模式；根据K个预测模式和权重导出模式，确定预测值。即本申请在确定K个模板时是基于当前块的大小和/或权重导出模式的，使得确定出的K个模板更加符合实际情况，这样使用这K个模板确定预测模式时，可以提高预测模式的确定准确性，进而使用准确确定的K个预测模式实现当前块的准确预测，提升编码效果。
上文以解码端为例对本申请的预测方法进行介绍，下面以编码端为例进行说明。
图21为本申请实一施例提供的预测方法流程示意图，本申请实施例应用于图1和图2所示视频编码器。如图21所示，本申请实施例的方法包括：
S201、确定当前块的权重导出模式。
需要说明的是，在本申请中，权重导出模式用于对当前块使用的权重进行确定。具体地，权重导出模式可以是导出权重的模式。对于一个给定长度和宽度的块，每一种权重导出模式可以导出一个权重矩阵；对于同样大小的块，不同权重导出模式导出的权重矩阵不同。
示例性的，在本申请中，AWP有56种权重导出模式，GPM有64种权重导出模式。
本申请中，编码端确定当前块的权重导出模式的方式包括但不限于如下几种：
方式一，上述权重导出模式为默认模式，例如，编码端默认权重导出模式为索引号为44的权重导出模式。
方式二，根据代价，确定出权重导出模式。例如，编码端尝试K个预测模式和权重导出模式的所有可能组合，K为大于1的正整数，选出代价最小的一个组合中的权重导出模式，确定为当前块的权重导出模式。
以K等于2为例，上述K个预测模式包括第一预测模式和第二预测模式，假设所有可用的预测模式有66种，第一预测模式有66种可能，由于第二预测模式与第一预测模式不相同，因此第二预测模式有65种，假设权重导出模式有63种(以GPM为例)，那么本申请可能使用任意两种不同的预测模式以及任意一种权重导出模式，共有66×65×63种可能。如果设定不使用PCM这种预测模式。那么就有65×64×63种可能。可见，在本申请中，还可以限制可以选择的预测模式，以及限制可以使用的权重导出模式的个数，那么组合的情况也会相应减少。
进一步地，在本申请的实施例中，编码端可以对所有可能组合进行代价计算，确定代价最小的一个组合。
假设K＝2，则每一个组合均为包括K个预测模式和一个权重导出模式。
可选地，为了减小代价计算的耗时，可以先对上述所有可能组合进行初选，如使用SAD，和SATD等作为近似的代价进行初选，确定设定数量的候选预测模式和权重导出模式的组合，再进行更详细的代价计算以实现细选，确定代价最小的一个预测模式和权重导出模式的组合。从而可以在初选时使用一些快速算法减少尝试的次数，比如说一个角度预测模式造成代价很大时，与它相邻的几个预测模式都不再尝试等。
在一些实施例中，编码端在确定当前块的权重导出模式之前，首先需要判断当前块是否使用K个不同的预测模式进行加权预测处理。若编码端确定当前块使用K个不同的预测模式进行加权预测处理时，则执行上述S201确定当前块的权重导出模式。
在一种可能的实现方式中，编码端可以通过确定当前块的预测模式参数，来确定当前块是否使用K个不同的预测模式进行加权预测处理。
可选的，在本申请的实施中，预测模式参数可以指示当前块是否可以使用GPM模式或AWP模式，即指示当前块是否可以使用K个不同的预测模式进行预测处理。
可以理解的是，在本申请的实施例中，可以将预测模式参数理解为一个表明是否使用了GPM模式或AWP模式标志位。具体地，编码端可以使用一个变量作为预测模式参数，从而可以通过对该变量的取值的设置来实现预测模式参数的设置。示例性的，在本申请中，如果当前块使用GPM模式或AWP模式，那么编码端可以将预测模式参数的取值设置为指示当前块使用GPM模式或AWP模式，具体地，编码端可以将变量的取值设置为1。示例性的，在本申请中，如果当前块不使用GPM模式或AWP模式，那么编码端可以将预测模式参数的取值设置为指示当前块不使用GPM模式或AWP模式，具体地，编码端可以将变量取值设置为0。进一步地，在本申请的实施例中，编码端在完成对预测模式参数的设置之后，便可以将预测模式参数写入码流中，传输至解码端，从而可以使解码端在解析码流之后获得预测模式参数。
需要说明的是，在本申请的实施例中，GPM模式或AWP模式为一种预测方法，具体地，为当前块确定K个不同的预测模式，然后分别根据这K个不同的预测模式确定出K个预测值，接着可以再确定权重，将则会K个预测值依据权重进行组合，最终便可以得到新的预测值。
需要说明的是，本申请实施例对上述K个不同的预测模式的具体类型不做限制。
图15为使用两种预测模式对当前块进行预测时的示意图，如图15所示，在对当前块进行预测时，可以使用第一预测模式确定第一预测值，同时使用第二预测模式确定第二预测值，然后可以利用权重对第一预测值和第二预测值进行组合处理，最终获得一个新的预测值。
在一些实施例中，在应用GPM模式或AWP模式时，可以对当前块的尺寸进行限制。
在本申请的一些实施例中，可以限制对一些大小的块不使用该GPM模式或AWP模式。因此，在本申请中，编码端可以先确定当前块的尺寸参数，然后根据尺寸参数确定当前块是否使用GPM模式或AWP模式。
在本申请的一些实施例中，还可以通过像素参数的限制来实现限制能够使用GPM模式或AWP模式的块的尺寸。
示例性的，在本申请中，可以有一个帧级的标志来确定当前待编码帧是否使用本申请。如可以配置帧内帧(如I帧)使用本申请，帧间帧(如B帧、P帧)不使用本申请。或者可以配置帧内帧不使用本申请，帧间帧使用本申请。或者可以配置某些帧间帧使用本申请，某些帧间帧不使用本申请。帧间帧也可以使用帧内预测，因而帧间帧也有可能使用本申请。
在一些实施例中，还可以有一个帧级以下、CU级以上(如tile、slice、patch、LCU等)的标志来确定这一区域是否使用本申请。
上述S201的具体实现过程可以参照上述S101的描述，在此不再赘述。
S202、根据当前块的大小和权重导出模式中的至少一个，确定K个模板。
其中，K为大于1的正整数。
本申请对当前块的模板的具体形状不做限制。
在一些实施例中，当前块的模板包括当前块的上方已编码区域和左侧已编码区域中的至少一个。
可选的，上方已编码区域的宽度与当前块的宽度相同，左侧已编码区域的高度与当前块的高度相同。
目前，对模板的划分不够精细，进而导致基于不精细的模板确定预测模式时，存在预测模式确定不准确，预测误差大的问题。
为了解决上述技术问题，本申请实施例通过当前块的大小和权重导出模式中的至少一个，来实现模板的精细划分。下面结合以下情况1和情况2所提出的方法，对上述S202中根据当前块的大小和权重导出模式中的至少一个，确定K个模板的过程进行详细介绍。
情况1，本申请实施例可以通过权重导出模式实现模板的更精细划分，具体的，上述S202包括如下步骤：
S202-A、根据权重导出模式，将当前块的模板划分为K个模板。
上述S202-A中根据权重导出模式，将当前块的模板划分为K个模板的方式包括但不限于如下几种：
方式一，根据权重导出模式对应的权重矩阵的分界线，将当前块的模板划分成K个模板。
例如图17A所示，本申请将当前块的权重导出模式对应的权重矩阵的分界线向当前块的模板中进行延长，以将模板进行划分，假设K＝2，则可以将分界线右侧的模板记为第一模板，将分界线左侧的模板记为第二模板。第一模板对应第一预测模式，第二模板对应的第二预测模式，在模板匹配时，可以使用第一模板导出第一预测模式，使用第二模板导出第二预测模式，进而实现预测模式的准确确定，提升编码效果。
在一些实施例中，根据上述方法划分的第一模板和第二模板可能不是长方形，例如图17A所示，第一模板和第二模板具有斜边，对不规则的模板计算代价较复杂。
为了降低模板匹配的复杂度，在一些实施例中，可以把第一模板和第二模板都划分成矩形。
在一些实施例中，若K大于2时，可以根据预设的划分方式，对根据权重导出模式划分后的模板再进行划分。假设K＝3，以图17B所示的根据权重导出模式划分后的模板为例，则可以将左侧模板划分成两部分，例如，将左侧模板的下半部分划分为第三模板，左侧模板的剩余上半部分和原有的上方模板的左半部分划分为第二模板，将上方模板的右半部分划分为第一模板，进而将当前块的模板划分成3个模板。
该方式一中，根据权重矩阵的分界线将当前块的模板划分成K个模板，该划分方式简单，且可以实现对模板的准确划分。
在一些实施例中，还可以根据如下方式二的方法将当前块的模板划分成K个模板。
方式二，上述S202-A包括如下S202-A1和S202-A2步骤：
S202-A1、将当前块的模板划分为M个子模板，M为大于或等于K的正整数；
S202-A2、根据权重导出模式，将M个子模板对应到K个模板中。
在该方式二中，首先将当前块的模板划分成多个子模板，例如划分成M个子模板，接着确定每个子模块对应到哪个模板，进而实现K个模板的划分。
本申请实施例对上述子模板的划分方式不做限制。
在方式二的一种可能实现方式1中，S202-A1包括：根据权重导出模式，将当前块的模板划分为M个子模板。
示例1，根据权重导出模式确定权重矩阵，将权重矩阵向当前块的模板延伸，例如向左向上延伸，将权重矩阵覆盖到当前块的模板上面。如图17D所示，当前块的模板包括当前块的左侧区域和上部区域，右下方的矩形区域为当前块。将当前块的权重矩阵向当前块的模板进行延伸，覆盖当前块的模板，这样可以根据当前块的模板被权重矩阵的覆盖情况，将当前块的模板划分成M个子模板。
示例性的，将图17D中黑色模板划分为第一子模板，将上侧灰色模板划分为第二子模板，将上侧白色模板划分为第二子模板。
示例性的，将图17D中左侧的黑色模板划分为第一子模板，将上侧的黑色模板划分为第二子模板，将上侧灰色模板划分为第三子模板，将上侧白色模板划分为第四子模板。
本申请对上述M个子模板的具体形状不做限制。
在一些实施例中，为了降低后续模式匹配计算的复杂度，则上述示例1将M个子模板划分为矩形。
示例2，根据权重导出模式，确定权重的分界线，将分界线向当前块的模板中进行延伸，以将当前块的模板划分为M个子模板。
具体的，根据权重导出模式确定权重的分界线，将该分界线向当前块的模板中进行延伸，将当前块的上方模板划分成两部分。这样可以根据权重分界线划分后的模板，确定M个子模板。
在一些实施例中，根据上述方法划分的第一模板和第二模板可能不是长方形，为了降低模板匹配的计算复杂度，在一些实施例中，将分界线向当前块的模板中进行延伸，得到分界线在当前块的模板中的延伸线；使用延伸线将当前块的模板划分为M个矩形子模板。例如图17G所示，使用延伸线将第一子模板和第二子模板划分为矩形。
该方式二中除了上述根据权重导出模式，将当前块的模板划分为M个子模板外，还可以采用下面实现方式2来将当前块的模板划分为M个子模板，具体如下所示。
在方式二的一种可能实现方式2中，根据预设的规则，将当前块的模板划分为M个子模板，即上述S202-A1包括如下步骤：
S202-A11、将当前块的上方模板划分成P个子模板；和/或，
S202-A12、将当前块的左侧模板划分成Q个子模板；
其中，P和Q均为小于或等于M的整数，且P与Q之和等于M。
本申请实施例中，当前块的模板包括当前块上方已编码的若干行像素行和当前块左侧已编码的若干列像素列，为了便于描述，本申请实施例将当前块的上方已编码的若干像素行记为当前块的上方模板，将当前块左侧已编码的若干列像素列记为当前块的左侧模板。在一些实施例中，当前块的模板还包括当前块的左上角已编码的区域，和/或包括当前块左下方的已编码区域等，本申请实施例对当前块的具体模板不做限制。本申请实施例主要对当前块的模板中的上方模板和左侧模板的划分为例进行说明。
在实现方式2中，将当前块的上方模板划分为P个子模板和/或将当前块的左侧模板划分为Q个子模板的方式不做限制，例如可以均等划分，或者按照预设的比例进行划分，或者按照预设的像素点数进行划分，或者按照预设的像素行数或像素列数进行划分等。
在一些实施例中，上述S202-A11中将当前块的左侧模板划分成P个子模板的方式包括但不限于如下几种：
方式1，沿着竖直方向，将上方模板划分为P个子模板。
在一种示例中，沿着竖直方向，将当前块的上方模板平均划分成P等份。
在另一种示例中，沿着竖直方向，按照预设的子模板比例，将当前块的上方模板划分为P个子模板。
方式2，根据预设的像素点数，将上方模板划分成P个子模板。
该方式2中，将预设的像素点数作为一个最小划分单元，对当前块的上方模板进行划分，划分为P个子模板。本申请对预设的像素点的具体排列方式不做限制。
在一些实施例中，将n列像素作为一个最小划分单元，将上方模板划分成P个子模板，n为正整数。
本申请对上述n的具体取值不做限制，例如为预设值。
可选的，当前块的上方模板的长度与当前块的长度相同，这样上述n可以根据当前块的长度确定，例如当前块的长度为n的正整数倍。示例性的，当前块的长度为16时，则该n可以为2、4、8等数值。
本申请实施例中，若需要对当前块的左侧模板进行划分时，则左侧模板的划分方式与上述当前块的上方模板的划分方式可以相同，也可以不同。
在一些实施例中，上述S202-A11中将当前块的左侧模板划分成Q个子模板的方式包括但不限于如下几种：
方式1，沿着水平方向，将左侧模板划分为Q个子模板。
在一种示例中，沿着水平方向，将当前块的左侧模板平均划分成Q等份。
在另一种示例中，沿着水平方向，按照预设的子模板比例，将当前块的左侧模板划分为Q个子模板。
方式2，根据预设的像素点数，将左侧模板划分成Q个子模板。
该方式2中，将预设的像素点数作为一个最小划分单元，对当前块的左侧模板进行划分，划分为Q个子模板。
在一些实施例中，将m行像素作为一个最小划分单元，将左侧模板划分成Q个子模板，m为正整数。
本申请对上述m的具体取值不做限制，例如为预设值。
可选的，当前块的左侧模板的宽度与当前块的宽度相同，这样上述m可以根据当前块的宽度确定，例如当前块的宽度为m的正整数倍。示例性的，当前块的宽度为16时，则该m可以为2、4、8等数值。
根据上述方式，将当前块的模板划分成M个子模板之后，执行上述S202-A2的步骤，即根据权重导出模式，将M个子模板对应到K个模板中。
在本申请实施例中，首先根据上述步骤，将当前块的模板划分成多个子模板，例如将当前块的模板划分为M个子模板，接着，确定这M个子模板中每个子模板属于对应到哪个模板，进而将M个子模板对应到K个模板中，实现对模板的精细准确划分。
上述S202-A2中根据权重导出模式，将M个子模板对应到K个模板中的实现方式包括但不限于如下几种：
方式1，根据权重矩阵的分界线，将M个子模板对应到K个模板中。
在一些实施例中，若权重的分界线将一个子模板划分成两部分，则可以将该子模板对应到第一模板和第二模板中，此时，第一模板和第二模板具有重叠的部分。
在一些实施例中，若权重的分界线将一个子模板划分为两个部分，则默认将该子模板对应到第一模板或第二模板。
在一些实施例中，权重的分界线将一个子模板划分为两个部分，若该子模板在第一预测模式中的区域大于在第二预测模式中的区域，则将该子模板对应到第一模板中。
方式2，根据子模板中像素点的权重，将M个子模板对应到K个模板中，具体的，上述S202-A2包括如下步骤：
S202-A21、对于M个子模板中的第j个子模板，根据权重导出模式，确定第j个子模板中的第一点关于第i个预测模式的权重，第i个预测模式为K个预测模式中的任一预测模式。
该方式2通过确定子模板中像素点的权重，确定是将该子模板划分为至哪个模板中，例如子模板中像素点的权重与第一预测模式对应的权重相同或基本相同，则将该子模板对应到第一模板中，若该子模板中像素点的权重与第二预测模式对应的权重相同或基本相同，则将该子模板对应到第二模板中。
由于确定M个子模板中每个子模板对应到哪个模板的过程相同，为了便于描述，本申请实施例以M个子模板中的第j个子模板为例进行说明，确定其他子模板对应到哪个模板的过程参照该第j个子模板即可。
在一些实施例中，可以根据第j个子模板中若干像素点的权重来判断是将该第j个子模板对应到哪个模板中。
在一些实施例中，为例降低计算复杂度，则通过确定第j个子模板中的一个像素点，例如第一点的权重，并根据该第一点的权重确定将第j个子模板对应到哪个模板中。
在一种示例中，上述第一点为第j个子模块中的任意一个点。
在一种示例中，上述第一点为第j个子模板与当前块的交界线上的一个点。例如为交界线上的任意一个点，或者为交界线的中点。
在一些实施例中，可以通过确定第j个子模板中的第一点关于K个预测模式中的任意一个预测模式的权重，来确定第j个子模板对应到哪个模板。
在一些实施例中，可以通过确定第j个子模板中的第一点分别关于K个预测模式权重，来确定第j个子模板对应到哪个模板。
其中确定第j个模板中的第一点关于K个预测模式中每个预测模式的权重的方式相同，本申请实施例以确定第一点关于第i个预测模式的权重为例进行说明。
上述S202-A21中确定第j个子模板中的第一点关于第i个预测模式的权重的方式包括但不限于如下示例：
在一种示例中，将当前块的权重矩阵向第j个子模板进行延伸，以使当前块的权重矩阵至少覆盖第j个子模板中的第一点，进而得到第一点的权重。
在另一种示例中，通过如下步骤S202-A211和S202-A212的步骤，确定第j个子模板中的第一点关于第i个预测模式的权重，即上述S202-A21包括如下步骤：
S202-A211、根据权重导出模式确定角度索引和距离索引；
S202-A212、根据角度索引和距离索引，确定第j个子模板中的第一点关于第i个预测模式的权重。
该实现方式中，通过权重导出模式，导出第j个子模板中的第一点关于第i个预测模式的权重，具体是，根据权 重导出模式，确定角度索引和距离索引，其中角度索引可以理解为权重导出模式导出的各权重的分界线角度索引。示例性的，可以根据上述表2，确定出权重导出模式对应的角度索引和距离索引，例如权重导出模式为27，则对应的角度索引为12，距离索引为3。接着，根据角度索引和距离索引，确定第j个子模板中的第一点关于第i个预测模式的权重。
在一些实施例中，上述S202-A212包括如下步骤：
S202-A2121、根据角度索引、距离索引和当前块的大小，确定第一点的第一参数；
S202-A2122、根据第一点的第一参数，确定第一点关于第i个预测模式的权重。
本实现方式中，根据角度索引、距离索引、模板的大小和当前块的大小，确定模板中各点的权重，进而将模板中每个点的权重组成的权重矩阵，确定为模板权重。
本申请的第一参数用于确定权重。在一些实施例中，第一参数也称为权重索引。
确定出第一点的第一参数weightIdx后，根据weightIdx确定出第一点(x，y)关于第i个预测模式的权重。
本申请中，上述S202-A2122中根据第一点的第一参数，确定第一点关于第i个预测模式的权重的方式包括的不限于如下几种：
一种方式，根据第一点的第一参数，确定第一点的第二参数；根据第一点的第二参数，确定第一点关于第i个预测模式的权重。
其中第二参数也用于确定权重。在一些实施例中，上述第二参数也称为第一分量下的权重索引，该第一分量可以为亮度分量、色度分量等。
例如，根据如下公式，确定第一点关于第i个预测模式的权重：
weightIdxL＝partFlip？32+weightIdx:32-weightIdx
wTemplateValue[x][y]＝Clip3(0,8,(weightIdxL+4)>>3)
其中，wTemplateValue[x][y]为第一点(x，y)关于第i个预测模式的权重，weightIdxL为第一点(x，y)的第二参数，partFlip为中间变量，根据角度索引angleIdx确定，例如上述所述：partFlip＝(angleIdx>＝13&&angleIdx<＝27)？0:1，也就是说，partFlip的值为1或0，当partFlip为0时，weightIdxL为32–weightIdx，当partFlip为1时，weightIdxL为32+weightIdx，需要说明的是，这里的32只是一种示例，本申请不局限于此。
另一种方式，根据第一点的第一参数、第一预设值和第二预设值，确定第一点关于第i个预测模式的权重。
为了降低第一点权重的计算复杂度，在一种方式中将第一点关于第i个预测模式的权重限定为第一预设值或第二预设值，也就是说，第一点关于第i个预测模式的权重要么为第一预设值，要么是第二预设值，进而降低第一点关于第i个预测模式的权重计算复杂度。
本申请对第一预设值和第二预设值的具体取值不做限制。
可选的，第一预设值为1。
可选的，第二预设值为0。
在一种示例，可以通过如下公式确定出第一点关于第i个预测模式的权重：
wTemplateValue[x][y]＝(partFlip？weightIdx:-weightIdx)>0？1:0
其中，wTemplateValue[x][y]为第一点(x，y)的权重，上述“1:0”中的1为第一预设值，0为第二预设值。
根据上述方法，确定出第j个子模板中的第一点关于第i个预测模式的权重后，执行如下S202-A22的步骤。
S202-A22、根据第j个子模板中的第一点关于第i个预测模式的权重，将第j个子模板对应到K个模板中。
该方式中，通过确定第j个子模板中的第一点关于第i个预测模式的权重，并根据第一点关于第i个预测模式的权重来确定是将该第j个子模板对应到哪个模板中。
在一种可能的实现方式中，若该第一点关于第i个预测模式的权重与该第i个预测模式的权重相同或基本相同时，则将该第j个子模板对应到第i个模板中。
在另一种可能的实现方式中，若第一点关于第i个预测模式的权重大于第一预设值，则将第j个子模板对应到第i个模板中，第i个模板为K个模板中的一个模板。例如，第j个子模板中的第一点关于第一预测模式的权重大于第一预设值，则将该第j个子模板对应到第一模板中。再例如，第j个子模板中的第一点关于第一预测模式的权重小于或等于第一预设值，则将该第j个子模板对应到第二模板中。
本申请对上述第一预测值的具体取值不做限制。
可选的，上述第一预设值为0。
可选的，上述第一预设值为小于权重中值的任意正数，若权重最大值为8，则权重中值为4。
在一些实施例中，若第一点关于第i个预测模式的权重大于第一预设值，且第一点关于第i+1个预测模式的权重也大于第一预设值，此时，可以将第j个子模板对应到第i个模板中，并将第j个子模板对应到第i+1个模板中，此时第i个模板与第i+1个模板具体重叠部分。以K＝2，第一预测值为0为例，假设第j个子模板为图18D中的子模板3，第一点为子模板3的下边中点，根据上述方法确定出第一点关于第一预测模式的权重大于0，且第一点关于第二预测模式的权重也大于0，此时，可以将子模板3对应到第一模板和第二模板中。
在一些实施例中，若K为2，i为1，则上述S202-A22包括如下几种示例：
示例1，若第一点关于第一预测模式的权重大于或等于第二预设值，则将第j个子模板对应到第一模板中。
示例2，若第一点关于第一预测模式的权重小于第二预设值，则将第j个子模板对应到第二模板中。
上述结合具体的示例，对情况1中根据权重导出模式，确定K个模板的具体实现方式进行介绍，例如根据权重导出模式对应的权重矩阵的分界线，将当前块的模板划分成K个模板，或者将当前块的模板划分为M个子模板，根据权重导出模式，将M个子模板对应到K个模板中。
本申请实施例，除了使用上述情况1的方法确定出K个模板外，还可以根据如下情况2的方式，确定出K个模板。
情况2，上述S202包括如下步骤：
S202-B1、从预设的不同块大小所对应的第一对应关系中，确定当前块对应的目标第一对应关系，第一对应关系包括不同角度索引或不同的权重导出模式与K个模板之间的对应关系；
S202-B2、从目标第一对应关系中，确定权重导出模式对应的K个模板。
由于当前块可能是正方形，也可能是长方形，可能长度比宽度大也可能宽度比长度大，而且比例也有1：2，1：4等可能。图10A和图10B示出了GPM在32x64块和64x32块的权重矩阵，可以看到不同形状下的划分线与块边界的交点并不一样。因为块形状变了但是划分线的角度并不根据块形状变化而变化。比如索引为52的模式，在32x64的块中与当前块左边界有交点，但是在64x32的块中与当前块左边界没有交点，而对应的交点在下边界。也就是说32x64的块中，模式52的黑色部分与当前块的左侧模板有相邻的部分，而在64x32的块中，模式52的黑色部分与当前块的左侧模板没有相邻的部分。
为了提高模板的选择准确性，本申请实施例根据当前块的长度和宽度设置不同的规则。
例如，对长度等于宽度，长度大于宽度，长度小于宽度这3种情况分别设置不同的第一对应关系，每一个第一对应关系可以为上述表5所示的表格，包括该情况下不同角度索引或不同的权重导出模式与K个模板之间的对应关系。
再例如，按照长宽比，如1：4，1：2，1：1，2：1，4：1等分类，为每一个分类设置一个第一对应关系，该第一对应关系包括该分类下不同角度索引或不同的权重导出模式与K个模板之间的对应关系。
这样编码时，编码端可以根据当前块的大小，例如当前块的长度和宽度，从预设的不同块大小所对应的第一对应关系中，确定当前块对应的目标第一对应关系，并根据权重导出模式，从该目标第一对应关系中，获得该权重导出模式对应的K个模板。在一些实施例中，若上述目标第一对应关系包括不同角度索引与K个模板之间的对应关系，则需要根据权重导出模式确定出目标角度索引，再根据目标角度索引从目标第一对应关系中查询该目标角度索引所对应的K个模板。
上述S202的具体实现过程可以参照上述S102的描述，在此不再赘述。
本申请实施例中，编码端根据上述步骤，确定出K个模板后，执行如下S203的步骤，根据K个模板确定当前块的K个预测模式。
S203、根据K个模板，确定K个预测模式。
在一些实施例中，上述S203包括如下S203-A1至S203-A4的步骤：
S203-A1、针对K个预测模式中的第i个预测模式，获取至少一个候选预测模式。
上述至少一个候选预测模式可以理解为第i个预测模式对应的候选预测模式，在一些实施例中，不同的预测模式对应的候选预测模式可以不同。在一些实施例中，若两个预测模式的类型相同时，例如均为帧内预测模式时，这两个预测模式对应的候选预测模式可以相同。
本申请实施例中，编码端在确定第i个预测模式时，首先判断该第i个预测模式是否是通过模板匹配方式确定的。
在一种可能的实现方式中，获取标志A，该标志A用于指示第i个预测模式是否通过模板匹配方式确定。
基于此，编码端判断该标志A的取值，若该标志A的取值为1时，则确定第i个预测模式为通过模板匹配方式确定的，此时，则编码端执行本申请实施例的方法，获取至少一个候选预测模式，并确定候选预测模式的代价，根据候选预测模式的代价，确定第j个预测模式。
在另一种可能的实现方式中，编码端默认该第i个预测模式是通过模板匹配方式确定的，这样编码端在确定第i个预测模式时，默认使用模板匹配方式确定第i个预测模式，接着获取至少一个候选预测模式，并确定候选预测模式的代价，根据候选预测模式的代价，确定第j个预测模式。
在一些实施例中，若上述第i个预测模式为帧间预测模式时，则上述至少一个候选预测模式包括一个或多个帧间预测模式，例如包括skip、merge、普通帧间预测模式、单向预测、双向预测、多假设预测等中的至少一个。
在一些实施例中，若上述第j个预测模式为帧内预测模式时，则上述至少一个候选预测模式包括直流(Direct Current，DC)模式、平面(PLANAR)模式、角度模式等中的至少一个。可选的，上述至少一个候选预测模式包括MPM列表中的帧内预测模式。
在一些实施例中，至少一个候选预测模式还可以包括IBC、palette等模式。
本申请对至少一个候选预测模式所包括的预测模式的类型以及预测模式的个数不做限制。
可选的，上述至少一个候选预测模式为预设模式。
可选的，上述至少一个候选预测模式为MPM列表中的模式。
可选的，上述至少一个候选预测模式是根据一些规则，如等间距筛选等，确定出的候选预测模式的集合。
S203-A2、使用候选预测模式对第i个模板进行预测，得到第i个模板的预测值；
示例性的，针对至少一个候选预测模式中的每一个候选预测模式，使用该候选预测模式对第i个模板进行预测，确定第i个模板的预测值。
其中，该第i个模板的预测值可以理解为由第i个模板中每个像素点的预测值组成的矩阵。
S203-A3、根据第i个模板的预测值和重建值，确定候选预测模式的代价。
示例性的，针对至少一个候选预测模式中的每一个候选预测模式，根据每个候选预测模式关于第i个模板的预测值，以及第i个模板的重建值，确定每个候选预测模式的代价。例如，根据候选预测模式关于第i个模板的预测值和第i个模板的重建值确定该候选预测模式对于第i个模板的损失，根据该候选预测模式对于第i个模板的损失，确定出候选预测模式的代价。
上述S203-A3中确定候选预测模式的代价的方式包括但不限于如下几种：
方式一，采用矩阵的方式确定候选预测模式的代价。
方式二，采用逐点计算的方式，确定候选预测模式的代价，即上述S203-A3包括如下步骤：
S203-A321、针对第i个模板中的第i个点，确定第i个点在第i个模板的预测值中对应的第i个预测值，与在第i样本的重建值中对应的第i个重建值之间的损失；
S203-A322、根据第i个点对应的损失，确定候选预测模式在第i个点处的代价；
S203-A323、根据候选预测模式在第i个模板中各点处的代价，确定候选预测模式的代价。
上述第i个点可以理解为第i个模板中的任意一个点，也就是说，确定第i个模板中每个点的代价的过程相同，参照第i个点即可。具体是，使用候选预测模式对第i个模板进行预测，得到候选预测模式关于第i个模板的预测值，将第i个点在第i个模板的预测值中对应的预测值记为第i个预测值，将该第i个点在第i个模板的重建值中对应的重建值记为第i个重建值，进而根据第i个预测值和第i个重建值，确定出该候选预测模式在第i个点的损失，并根据该候选预测模式在第i个点的损失，确定候选预测模式在第i个点处的代价，例如将该候选预测模式在第i个点的损失确定为该候选预测模式在第i个点处的代价。根据上述方法，确定出该候选预测模式在第i个模板中的每个点或多个点处的代价，进而根据第i个模板中每个点或多个点的代价，确定出候选预测模式关于第i个模板的代价。例如，将候选预测模式在第i个模板中各点处的代价之和，确定为候选预测模式关于第i个模板的代价，或者将候选预测模式在第i个模板中各点处的代价平均值，确定为候选预测模式关于第i个模板的代价，本申请对根据第i个模板中至少一个点的代价，确定候选预测模式关于第i个模板的代价不做限制。
示例性的，以SAD代价为例，可以根据如下公式(3)确定出候选预测模式在第i个模板中的第i个点(x，y)处的代价：
tempValueA[x][y]＝abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y])(3)
示例性的，根据如下公式(4)确定出候选预测模式的代价：
costCandA＝∑tempValueA[x][y]    (4)
其中，abs(predTemplateSamplesCandA[x][y]-recTemplateSamples[x][y])是第i个模板中点(x，y)的预测值predTemplateSamplesCandA和重建值recTemplateSamples的差的绝对值，将该差的绝对值称为点(x，y)对应的损失。tempValueA[x][y]可以认为是该候选预测模式在这个点(x，y)的代价。该候选预测模式在第i个模板上的总的代价costCandA为第i个模板上每一个点的代价累加。
需要说明的是，上述以SAD为例来确定候选预测模式的代价，可选的还可以根据SATD、MSE等代价计算方法，确定候选预测模式关于第i个模板的代价。
根据上述方法，可以确定出候选预测模式关于第i个模板的代价，接着执行如下S203-A4的步骤。
S203-A4、根据至少一个候选预测模式的代价，确定第i个预测模式。
本申请实施例，若第i个预测模式为通过模板匹配方法确定，则通过上述方法，确定出候选预测模式的代价，并根据各候选预测模式的代价，确定第i个预测模式。
示例1，将至少一个候选预测模式中代价最小的候选预测模式，确定为第i个预测模式。
示例2，根据候选预测模式的代价，从至少一个候选预测模式中选出一个或多个候选预测模式；根据一个或多个候选预测模式，确定第j个预测模式。
在该示例2的一种可能的实现方式中，编码端从一个或多个候选预测模式中选出一个候选预测模式，作为第j个预测模式。
例如，上述一个或多个候选预测模式为M个，编码端将这M个候选预测模式按照代价进行排序，例如按照代价从小到大对M个候选预测模式进行排序，或者按照代价从大到小对M个候选预测模式进行排序，从排序后的M个候选预测模式中确定出一个候选预测模式B作为第i个预测模式。同时，编码端将该候选预测模式模式B的标识编入码流中，该候选预测模式B的标识可以是候选预测模式B在M个候选预测模式中的排序号，也可以是该候选预测模式B的模式索引号。这样，解码端通过解码码流，得到该候选预测模式B的标识，进而根据候选预测模式B的标识，将上述确定的M个候选预测模式中候选预测模式B的标识对应的候选预测模式，确定为第i个预测模式。
在该示例2的另一种可能的实现方式中，编码端获取当前块的备选预测模式；确定备选预测模式对第i个模板进行预测时的代价；根据备选预测模式对第i个模板进行预测时的代价以及上述选出的一个或多个候选预测模式关于第i个模板的代价，从备选预测模式和上述一个或多个候选预测模式中选出一个预测模式，作为第i个预测模式。
可选的，上述当前块的备选预测模式包括当前块周围已重建编码块的预测模式和/或预设预测模式中的一个或多个。
可以理解的是，在本申请中，预设预测模式可以包括DC模式、Bilinear模式、Planar模式等多种不同模式中的一种或多种。
具体的，解码端获取当前块的备选预测模式，例如将当前块周围已重建解码块的预测模式和/或预设预测模式中的一个或多个作为当前块的备选预测模式。接着，确定每个备选预测模式对模板进行预测时的代价，例如使用备选预测模式对当前块进行预测，得到预测值，将该预测值与模板的重建值进行比较，得到该备选预测模式的代价，其中备选预测模式的代价可以是SAD、SATD等代价。根据备选预测模式的代价以及上述一个或多个候选预测模式的代价，从备选预测模式和上述一个或多个候选预测模式中选出一个预测模式作为第j个预测模式，例如将备选预测模式和上述一个或多个候选预测模式中代价最小的预测模式，确定为第j个预测模式。
值得注意的是，上述当前块的备选预测模式与上述确定的一个或多个候选预测模式不相同，也就是说，解码端将当前块周围已重建解码块的预测模式和/或预设预测模式中与上述一个或多个候选预测模式中相同的预测模式删除，将剩余的预测模式确定为当前块的备选预测模式。
可以理解的是，对帧间预测来说，模板匹配可以在一个初始运动信息的基础上进行“搜索”。一个预测模式需要确定一个运动信息。可以在一个初始运动信息的周边一定范围内确定一些运动信息，从而确定一些预测模式。如给定一个初始运动信息，其运动矢量是(xInit，yInit)，设置一个搜索范围如水平方向从xInit-sR到xInit+sR，竖直方向从yInit-sR 到yInit+sR的矩形区域，其中sR可以是2，4，8等。该矩形区域内的每一个运动矢量都可以和初始运动信息的其他信息，如参考帧索引和预测列表标志等组合起来确定一个运动信息，从而确定一个预测模式。上述至少一个候选预测模式可以包含所述确定的预测模式。比如如果GPM在merge模式下使用，如果使用模板匹配的方法确定第一预测模式，可以使用merge_gpm_idx0从mergeCandList中确定一个初始运动信息。再根据上述方法确定(2*sR+1)*(2*sR+1)个运动信息，从而确定一些预测模式，这些预测模式都是merge模式，或者称为使用模板匹配的merge模式。
值得注意的是，如果上述至少一个候选预测模式组成的集合包含的预测模式很多，出于复杂度的考虑，可能不会对至少一个候选预测模式中的每一个候选预测模式都确定其代价。在一些实施例中，根据至少一个候选预测模式的代价，确定第j个预测模式的过程也可以进一步扩展为几层的由粗选到细选的过程。比如帧间预测模式中，运动矢量支持分像素精度，如1/4、1/8、1/16精度等。所以可以先从包含整像素运动矢量的预测模式中选出代价最小的预测模式，再在该预测模式和运动矢量在该模式的运动矢量附近的包含分像素运动矢量的预测模式中进一步选出代价最小的预测模式。比如在帧内预测模式中，根据候选预测模式的代价，先按一定粒度选出一个或几个帧内预测模式，再在该一个或几个帧内预测模式及更细粒度的相邻帧内预测模式中再进行筛选。
本申请实施例中，若K个预测模式中的第i个预测模式通过模板匹配方式确定时，通过获取至少一个候选预测模式，并使用候选预测模式对模板进行预测，得到该候选预测模式下模板的预测值；根据该候选预测模式下模板的预测值和模板的重建值，得到该候选预测模式的代价，最后根据候选预测模式的代价，得到第j个预测模式。
上述实施例以K个预测模式中的第i个预测模式的确定过程为例进行说明，K个预测模式的中的其他预测模式的确定过程与上述第i个预测模式的确定过程一致，参照即可。例如，K＝2，根据上述方法确定出第一预测模式和第二预测模式，接着，编码端根据该第一预测模式得到第一预测值，根据第二预测模式得到第二预测值，将第一预测值和第二预测值进行加权，得到新的预测值。
上述S203的具体实现过程可以参照上述S103的描述，在此不再赘述。
本申请实施例根据上述方法，可以根据K个模板确定出K个预测模式，接着使用这K个预测模式对当前块进行预测，得到当前块的预测值，具体参照下面的S204所述。
S204、根据K个预测模式和权重导出模式，确定预测值。
本申请，根据权重导出模式确定权重，根据K个预测模式确定K个预测值，根据权重对K个预测值进行加权，将加权结果确定为最终的预测值。
在本申请中，权重导出模式用于确定当前块的预测值进行加权时的权重。具体地，权重导出模式可以是导出权重的模式。对于一个给定长度和宽度的块，每一种权重导出模式可以导出一个权重矩阵；对于同样大小的块，不同权重导出模式导出的权重矩阵可以不同。
示例性的，在本申请中，AVS3的AWP有56种权重导出模式，VVC的GPM有64种权重导出模式。
可以理解的是，在本申请的实施例中，编码端在基于K个预测模式以及权重，确定预测值时，可以先根据K个预测模式中的每个预测模式，确定每个预测模式对应的预测值，对每个预测模式对应的预测值进行加权，得到最终的预测值。
在一些实施例中，上述预测过程是以像素点为单位进行的，对应的上述权重也为像素点对应的权重。此时，对当前块进行预测时，使用K个预测模式中的每个预测模式对当前块中的某一个像素点A进行预测，得到K个预测模式关于像素点A的K个预测值，根据像素点A的权重对这K个预测值进行加权，得到像素点A的最终预测值。对当前块中的每一个像素点执行上述步骤，可以得到当前块中每个像素点的最终预测值，当前块中每个像素点的最终预测值构成当前块的最终预测值。以K＝2为例，使用第一预测模式对当前块中的某一个像素点A进行预测，得到该像素点A的第一预测值，使用第二预测模式对该像素点A进行预测，得到该像素点A的第二预测值，根据像素点A对应的权重，对第一预测值和第二预测值进行加权，得到像素点A的最终预测值。
在一种示例中，以K＝2为例，若第一预测模式和第二预测模式均为帧内预测模式时，采用第一帧内预测模式进行预测，得到第一预测值，采用第二帧内预测模式进行预测，得到第二预测值，根据预测权重对第一预测值和第二预测值进行加权，得到新的预测值。例如，采用第一帧内预测模式对像素点A进行预测，得到像素点A的第一预测值，采用第二帧内预测模式对像素点A进行预测，得到像素点A的第二预测值，根据像素点A对应的预测权重，对第一预测值和第二预测值进行加权，得到像素点A的最终预测值。
在一些实施例中，若K个预测模式中的第i个预测模式为帧间预测模式，则上述根据K个预测模式和权重导出模式，确定预测值包括如下步骤：
S204-AB21、根据第i个预测模式，确定运动信息；
S204-AB22、根据运动信息，确定第i个预测值；
S204-AB23、根据K个预测模式中除第i个预测模式外的其他预测模式，确定K-1个预测值；
S204-AB24、根据权重导出模式，确定权重；
S204-AB25、根据第i个预测值、K-1个预测值和权重，确定预测值。
以K＝2为例，若第一预测模式为帧内预测模式，第二预测模式为帧间预测模式时，采用帧内预测模式进行预测，得到第一预测值，采用帧间预测模式进行预测，得到第二预测值，根据预测权重对第一预测值和第二预测值进行加权，得到新的预测值。在该示例中，采用帧内预测模式对当前块中每一个点进行预测，得到当前块中每一个点的预测值，当前块中每一个点的预测值，构成当前块的第一预测值。采用帧间预测模式，确定一运动信息，根据该运动信息确定当前块的最佳匹配块，将该最佳匹配块确定为当前块的第二预测值。针对当前块中每个像素点的预测权重，对当前块的第一预测值和第二预测值进行逐点加权运算，得到当前块的新的预测值。例如，对于当前块中的像素点A，根据像素点A的预测权重，将当前块的第一预测值中像素点A对应的第一预测值，与当前块的第二预测值中像素点A对应的第二预测值进行加权，得到像素点A的最终预测值。
在一些实施例中，在执行本申请实施例的方法之前，编码端需要判断当前块是否适用于模板匹配方法，若编码端 确定当前块适用于模板匹配方法，则执行上述S201至S204的步骤，若编码端确定当前块不适用于模板匹配方法时，则使用其他的方式确定K个预测模式。
在一些实施例中，编码端根据K个模板所包括的点数，确定当前块是否适用于模板匹配方法。
本申请实施例中，可用模板比较大的预测模式使用模板匹配或相邻重建像素的纹理特性，对可用模板比较小的预测模式则不使用模板匹配或相邻重建像素的纹理特性。
在一种可能的实现方式中，可以根据K个模板所包括的点数，确定当前块是否适用于模板匹配方法。
在一种示例中，若K个模板所包括的点数均大于预设阈值时，确定当前块适用于模板匹配方法，进而执行上述S203根据K个模板，确定K个预测模式。
可选的，上述预设阈值可以是0。
可选的，上述预设阈值为权重中值，例如4。
可选的，上述预设阈值为一个定值。
可选的，上述预设阈值根据当前块的尺寸确定，例如为当前块的总点数的1/m1，m1为正数。
可选的，上述预设阈值根据当前块的模板的尺寸确定，例如为当前块的模板的总点数的1/m2，m2为正数。
在另一种示例中，若K个模板中至少一个模板包括的点数小于预设阈值时，则根据权重导出模式，确定K个预测模式。
本申请实施例中，编码端根据上述S202的步骤，基于当前块的大小和权重导出模式中的至少一个，确定K个模板后，根据K个模板所包括的点数，确定当前块是否适用于模板匹配方法。具体是，针对K个模板中的第i个模板，若该第i个模板所包括的像素点的个数大于预设阈值时，则说明第i个模板中用于确定第i个预测模式的可用模板较大，使用该第i个模板确定第i个预测模式时，可以提高预测效果。若该第i个模板所包括的像素点的个数小于预设阈值时，则说明该第i个模板中用于确定第i个预测模式的可用模板较小或不存在，此时使用模板匹配方法确定第i个预测模式时，不仅不会提高压缩效率反而可能起到反作用。
在一些实施例中，编码端将第一标志写入码流，该第一标志用于指示当前块是否采用模板匹配方式导出预测模式，若编码端确定当前块采用模板匹配方式导出预测模式时，则将该第一标志置为1，且将置为1的第一标志写入码流，若编码端确定当前块不采用模板匹配方式导出预测模式时，则将该第一标志置为0，且将置为0的第一标志写入码流。这样，解码端获得码流后，通过解码码流，得到该第一标志，并且根据该第一标志确定当前块是否采用模板匹配方式导出预测模式。
在一些实施例中，若确定当前块不适用于模板匹配方法时，则根据权重导出模式，确定K个预测模式中的至少一个。
在本申请中，权重变化的位置构成一条直线(曲线段)，或，如图4和图5所示过渡区域中权重相同的位置构成一条直线(曲线段)。可以将这条直线叫做分界线(或划分线或分割线)。分界线本身也是有角度的，可以设水平向右的角度为0，角度逆时针增加。那么分界线可能有水平0度，竖直90度，倾斜的如45度，135度，以及其他各种不同的角度等。如果一个块选择使用某一个权重矩阵，那么对应的纹理很可能在分界线两边显现出不同的特性，比如分界线两边是两种不同角度的纹理，或者，分界线的一边是一种角度的纹理，而另一边是一种比较平坦的纹理。由于分界线本身也是有角度的，因此可以假设一个点经过角度预测得到的，它可能与当前块的某些纹理是接近的，因而这条直线与当前块的两个预测模式是存在相关性的。
具体地，在本申请中，假设分界线是由一个点经过角度预测得到的，那么可以找到至少一个角度预测模式，这个角度预测模式可以近似地做出分界线。
需要说明的是，在本申请中，权重导出模式也可以为权重的索引，如AWP的56种模式就可以认为是56种权重导出模式，VVC的GPM的64种模式就可以认为是64种权重导出模式。
在一些实施例中，除了与权重分界线对应的帧内角度预测模式使用的可能性高之外，和它相关的某些帧内角度预测模式使用的可能性也较高。比如说与这个分界线相近的角度，或这个分界线垂直的角度等对应的帧内角度预测模式。
在一些实施例中，如果GPM的一个预测值来自于帧内预测，一个预测值来自于帧间预测。若本申请所使用的帧内预测模式默认由权重导出模式确定。比如权重导出模式的分界线是水平方向的，如图4中GPM的索引为18，19，50，51的模式，就确定帧内预测模式为水平方向的模式18。比如权重导出模式的分界线是竖直方向的，如图4中GPM的索引为0，1，36，37的模式，就确定帧内预测模式为垂直方向的模式50。
也就是说，本申请在根据权重导出模式确定K个预测模式中的至少一个之前，首先要确定K预测模式的类型，当预测模式为帧内预测模式时，才可以根据权重导出模式确定该预测模式。
在此基础上，在根据权重导出模式确定K预测模式中的至少一个之前，本申请实施例的方法还包括：
步骤21-0、获取类型标志，该类型标志用于指示K预测模式分别是否属于帧内预测模式；
步骤21-1、根据类型标志，确定K个预测模式的类型。
示例性的，以K＝2为例，若类型标志的取值为第一数值时，指示第一预测模式和第二预测模式均为帧间预测模式，此时，mode0IsInter为1，mode1IsInter为1，其中mode0IsInter指示第一预测模式是否为帧间预测模式，mode1IsInter指示第二预测模式是否为帧间预测模式，当第一预测模式为帧间预测模式时，mode0IsInter为1，当第二预测模式为帧间预测模式时，mode1IsInter为1。
示例性的，类型标志的取值为第二数值时，指示第一预测模式为帧内预测模式，第二预测模式为帧间预测模式，此时，mode0IsInter为0，mode1IsInter为1。
示例性的，类型标志的取值为第三数值时，指示第一预测模式为帧间预测模式，第二预测模式为帧内预测模式，此时，mode0IsInter为1，mode1IsInter为0。
示例性的，类型标志的取值为第四数值时，指示第一预测模式和第二预测模式均为帧内预测模式，此时， mode0IsInter为0，mode1IsInter为0。
本申请对上述第一数值、第二数值、第三数值和第四数值的具体取值不做限制。
可选的，第一数值为0。
可选的，第二数值为1。
可选的，第三数值为2。
可选的，第四数值为3。
在一种示例中，可以用字段intra_mode_idx表示类型标志。
本申请中，编码端根据类型标志确定出第一预测模式和第二预测模式的类型后，在编码时，需要将该类型标志编入码流，以使根据类型标志确定第一预测模式和第二预测模式的类型。
本申请中，编码端根据上述类型标志，确定出K预测模式的类型后，若K预测模式中至少一个为帧内预测模式时，基于权重导出模式确定该帧内预测模式。
也就是说，本申请中基于权重导出模式来确定帧内预测模式，例如，第一预测模式和第二预测模式均为帧内预测模式时，则基于权重导出模式来确定第一预测模式和第二预测模式。再例如，第一预测模式和第二预测模式中的一个为帧内预测模式时，则基于权重导出模式确定第一预测模式和第二预测模式中的帧内预测模式。
本申请中，基于权重导出模式确定K个预测模式中的至少一个的方式包括但不限于如下几种：
方式一，若K个预测模式中的至少一个为帧内预测模式时，则根据权重导出模式确定角度索引，并将角度索引对应的帧内预测模式，确定为K个预测模式中的一个预测模式。
在一些实施例中，用字段angleIdx表示角度索引。
上述表2示出了merge_gpm_partition_idx与angleIdx的对应关系，参照上述表2，可以根据该权重导出模式导出角度索引。
本申请中，角度索引与帧内预测模式具有对应关系，即不同的角度索引与不同帧内预测模式对应。
示例性的，角度索引与帧内预测模式具有对应关系如上述表7所示。
该方式一中，以K＝2为例，若第一预测模式或第二预测模式为帧内预测模式时，则根据权重导出模式确定角度索引。接着，在上述表7中，确定出该角度索引对应的帧内预测模式，例如，角度索引为2，其对应的帧内预测模式为42，进而将帧内预测模式42确定为第一预测模式或第二预测模式。
方式二，若K个预测模式中的至少一个为帧内预测模式时，则获取权重导出模式对应的帧内预测模式；根据权重导出模式对应的帧内预测模式，确定K个预测模式中的至少一个。
该方式二中，以K＝2为例，若第一预测模式和/或第二预测模式为帧内预测模式时，第一预测模式和/或第二预测模式是从权重导出模式对应的帧内预测模式中确定出。例如，第一预测模式和/或第二预测模式可以是跟权重划分线(也称为分界线)同一条直线的或近似同一条直线的帧内预测模式。或者，第一预测模式和/或第二预测模式可以是跟权重划分线垂直的或近似垂直的帧内预测模式。
由上述可知，与权重导出模式对应的帧内预测模式的类型较多，例如包括与权重的分界线平行的帧内预测模式、与分界线垂直的帧内预测模式等。本申请可以通过标志来指示第一预测模式和/或第二预测模式具体选择权重导出模式对应的帧内预测模式中的那一个模式。
示例性的，以K＝2为例，若第一预测模式为帧内预测模式，则使用第二标志来指示第一预测模式与权重导出模式对应的帧内预测模式的对应关系，例如第二标志指示第一预测模式为与权重的分界线平行的帧内预测模式，或者指示第一预测模式为与权重的分界线垂直的帧内预测模式。
示例性的，若第二预测模式为帧内预测模式，则使用第三标志来指示第二预测模式与权重导出模式对应的帧内预测模式的对应关系，例如第三标志指示第二预测模式为与权重的分界线平行的帧内预测模式，或者指示第二预测模式为与权重的分界线垂直的帧内预测模式。
基于此，上述方式二中根据权重导出模式对应的帧内预测模式，确定第一预测模式和/或第二预测模式的方式包括但不限于如下几种示例：
示例1，若第一预测模式为帧内预测模式时，则获取第二标志，并将权重导出模式对应的帧内预测模式中第二标志对应的帧内预测模式，确定为第一预测模式。
示例2，若第二预测模式为帧内预测模式时，则获取第三标志，并将权重导出模式对应的帧内预测模式中第三标志对应的帧内预测模式，确定为第二预测模式。
在一些实施例中，权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式和与分界线垂直的帧内预测模式中的至少一个。
在一些实施例中，权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式、与分界线垂直的帧内预测模式和planar模式中的至少一个。
示例性的，可以根据如上述表8的方式，将第二标志(intra_gpm_idx0)和/或第三标志(intra_gpm_idx1)写入码流。
编码端通过上述表8的方式，将第二标志和/或第三标志写入码流。解码端解码码流，得到第二标志和/或第三标志，并根据该第二标志和/或第三标志，确定第一预测模式和/或第二预测模式，进而使用第一预测模式和第二预测模式以及权重，确定预测值。
参照上述方法，根据权重导出模式确定K个预测模式中的至少一个，根据K个预测模式确定K个预设值，对这K个预测值进行加权处理，得到最终的预测值。
上述S204的具体实现过程可以参照上述S104的描述，在此不再赘述。
本申请实施例提供的预测方法，编码端通过确定当前块的权重导出模式；根据当前块的大小和权重导出模式中的至少一个，确定K个模板；根据K个模板，确定K个预测模式；根据K个预测模式和权重导出模式，确定预测值。 即本申请在确定K个模板时是基于当前块的大小和/或权重导出模式的，使得确定出的K个模板更加符合实际情况，这样使用这K个模板确定预测模式时，可以提高预测模式的确定准确性，进而使用准确确定的K个预测模式实现当前块的准确预测，提升编码效果。
应理解，图14至图21仅为本申请的示例，不应理解为对本申请的限制。
以上结合附图详细描述了本申请的优选实施方式，但是，本申请并不限于上述实施方式中的具体细节，在本申请的技术构思范围内，可以对本申请的技术方案进行多种简单变型，这些简单变型均属于本申请的保护范围。例如，在上述具体实施方式中所描述的各个具体技术特征，在不矛盾的情况下，可以通过任何合适的方式进行组合，为了避免不必要的重复，本申请对各种可能的组合方式不再另行说明。又例如，本申请的各种不同的实施方式之间也可以进行任意组合，只要其不违背本申请的思想，其同样应当视为本申请所公开的内容。
还应理解，在本申请的各种方法实施例中，上述各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。另外，本申请实施例中，术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系。具体地，A和/或B可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本申请中字符“/”，一般表示前后关联对象是一种“或”的关系。
上文结合图14至图21，详细描述了本申请的方法实施例，下文结合图22至图25，详细描述本申请的装置实施例。
图22是本申请一实施例提供的预测装置的示意性框图，该预测装置10应用于上述视频解码器。
如图20所示，预测装置10包括：
解码单元11，用于解码码流，确定当前块的权重导出模式；
模板确定单元12，用于根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
模式确定单元13，用于根据所述K个模板，确定K个预测模式；
预测单元14，用于根据所述K个预测模式和所述权重导出模式，确定预测值。
在一些实施例中，模板确定单元12，具体用于根据所述权重导出模式，将所述当前块的模板划分为所述K个模板。
在一些实施例中，模板确定单元12，具体用于将所述当前块的模板划分为M个子模板，所述M为大于或等于K的正整数；根据所述权重导出模式，将所述M个子模板对应到所述K个模板中。
在一些实施例中，模板确定单元12，具体用于根据所述权重导出模式，将所述当前块的模板划分为M个子模板。
在一些实施例中，模板确定单元12，具体用于根据所述权重导出模式，确定权重的分界线；将所述分界线向所述当前块的模板中进行延伸，以将所述当前块的模板划分为M个子模板。
在一些实施例中，模板确定单元12，具体用于将所述分界线向所述当前块的模板中进行延伸，得到所述分界线在所述当前块的模板中的延伸线；使用所述延伸线，将所述当前块的模板划分为M个矩形子模板。
在一些实施例中，模板确定单元12，具体用于将所述当前块的上方模板划分成P个子模板；和/或，将所述当前块的左侧模板划分成Q个子模板；其中，所述P和Q均为小于或等于M的整数，且所述P与Q之和等于M。
在一些实施例中，模板确定单元12，具体用于沿着竖直方向，将所述上方模板划分为P个子模板。
在一些实施例中，模板确定单元12，具体用于根据预设的像素点数，将所述上方模板划分成P个子模板。
在一些实施例中，模板确定单元12，具体用于将n列像素作为一个最小划分单元，将所述上方模板划分成P个子模板，所述n为正整数。
可选的，所述n根据所述当前块的长度确定。
在一些实施例中，模板确定单元12，具体用于沿着水平方向，将所述左侧模板划分为Q个子模板。
在一些实施例中，模板确定单元12，具体用于根据预设的像素点数，将所述左侧模板划分成Q个子模板。
在一些实施例中，模板确定单元12，具体用于将m行像素作为一个最小划分单元，将所述左侧模板划分成Q个子模板，所述m为正整数。
可选的，所述m根据所述当前块的宽度确定。
在一些实施例中，模板确定单元12，具体用于对于所述M个子模板中的第j个子模板，根据所述权重导出模式，确定所述第j个子模板中的第一点关于第i个预测模式的权重，所述第i个预测模式为所述K个预测模式中的任一预测模式；根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中。
在一些实施例中，模板确定单元12，具体用于根据所述权重导出模式确定角度索引和距离索引；根据所述角度索引和距离索引，确定所述第j个子模板中的第一点关于第i个预测模式的权重。
在一些实施例中，所述第一点为所述第j个子模板与所述当前块的交界线上的一个点。
可选的，所述第一点为所述交界线的中点。
在一些实施例中，模板确定单元12，具体用于根据所述角度索引、所述距离索引和所述当前块的大小，确定所述第一点的第一参数，所述第一参数用于确定权重；根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重。
在一些实施例中，模板确定单元12，具体用于根据所述第一点的第一参数，确定所述第一点的第二参数；根据所述第一点的第二参数，确定所述第一点关于第i个预测模式的权重。
在一些实施例中，模板确定单元12，具体用于根据所述第一点的第一参数、第一预设值和第二预设值，确定所述第一点关于第i个预测模式的权重。
在一些实施例中，所述第一点关于第i个预测模式的权重为第一数值或为第二数值。
在一些实施例中，模板确定单元12，具体用于若所述第一点关于第i个预测模式的权重大于第一预设值，则将所述第j个子模板对应到第i个模板中，所述第i个模板为所述K个模板中的一个模板。
在一些实施例中，若所述K为2，i为1，模板确定单元12，具体用于若所述第一点关于第一预测模式的权重大于 或等于第二预设值，则将所述第j个子模板对应到第一模板中；若所述第一点关于第一预测模式的权重小于第二预设值，则将所述第j个子模板对应到第二模板中。
在一些实施例中，模板确定单元12，具体用于从预设的不同块大小所对应的第一对应关系中，确定所述当前块对应的目标第一对应关系，所述第一对应关系包括不同角度索引或不同的权重导出模式与所述K个模板之间的对应关系；从所述目标第一对应关系中，确定所述权重导出模式对应的所述K个模板。
在一些实施例中，模式确定单元13，具体用于针对所述K个预测模式中的第i个预测模式，获取至少一个候选预测模式；使用所述候选预测模式对所述K个模板中的第i个模板进行预测，得到所述第i个模板的预测样本；根据所述第i个模板的预测值和重建值，确定所述候选预测模式的代价；根据所述至少一个候选预测模式的代价，确定所述第i个预测模式。
在一些实施例中，预测单元14，具体用于根据所述权重导出模式，确定权重；根据所述K个预测模式，确定K个预测值；根据所述权重对所述K个预测值进行加权，得到最终的预测值。
在一些实施例中，模式确定单元13，具体用于若所述K个模板所包括的点数均大于预设阈值时，则根据所述K个模板，确定K个预测模式。
在一些实施例中，模式确定单元13，还用于若所述K个模板中至少一个模板包括的点数小于预设阈值时，则根据所述权重导出模式，确定所述K个预测模式。
在一些实施例中，模板确定单元12，具体用于解码所述码流，得到第一标志，所述第一标志用于指示是否采用模板匹配方式导出预测模式；若所述第一标志指示采用模板匹配方式导出预测模式时，则根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板。
在一些实施例中，模式确定单元13，还用于若所述第一标志指示不采用模板匹配方式导出预测模式时，则根据所述权重导出模式，确定所述K个预测模式。
在一些实施例中，所述若所述K个预测模式中的至少一个为帧内预测模式时，则模式确定单元13，具体用于根据所述权重导出模式确定角度索引；将所述角度索引对应的帧内预测模式，确定为所述K个预测模式中的至少一个。
在一些实施例中，所述若所述K个预测模式中的至少一个为帧内预测模式时，模式确定单元13，具体用于获取所述权重导出模式对应的帧内预测模式；根据所述权重导出模式对应的帧内预测模式，确定所述K个预测模式中的至少一个。
在一些实施例中，所述权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式、与所述分界线垂直的帧内预测模式和planar模式中的至少一个。
在一些实施例中，若所述K个预测模式中的第i个预测模式为帧间预测模式时，则上述预测单元14，具体用于根据所述第i个预测模式，确定运动信息；根据所述运动信息，确定第i个预测值；根据所述K个预测模式中除所述第i个预测模式外的其他预测模式，确定K-1个预测值；根据所述权重导出模式确定权重；根据所述第i个预测值、所述K-1个预测值和所述权重，确定最终的预测值。
应理解，装置实施例与方法实施例可以相互对应，类似的描述可以参照方法实施例。为避免重复，此处不再赘述。具体地，图22所示的装置10可以执行本申请实施例的解码端的预测方法，并且装置10中的各个单元的前述和其它操作和/或功能分别为了实现上述解码端的预测方法等各个方法中的相应流程，为了简洁，在此不再赘述。
图23是本申请一实施例提供的预测装置的示意性框图，该预测装置应用于上述编码器。
如图23所示，该预测装置20可以包括：
确定单元21，用于确定当前块的权重导出模式；
模板确定单元22，用于根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
模式确定单元23，用于根据所述K个模板，确定K个预测模式；
预测单元24，用于根据所述K个预测模式和所述权重导出模式，确定预测值。
在一些实施例中，模板确定单元22，具体用于根据所述权重导出模式，将所述当前块的模板划分为所述K个模板。
在一些实施例中，模板确定单元22，具体用于将所述当前块的模板划分为M个子模板，所述M为大于或等于K的正整数；根据所述权重导出模式，将所述M个子模板对应到所述K个模板中。
在一些实施例中，模板确定单元22，具体用于根据所述权重导出模式，将所述当前块的模板划分为M个子模板。
在一些实施例中，模板确定单元22，具体用于根据所述权重导出模式，确定权重的分界线；将所述分界线向所述当前块的模板中进行延伸，以将所述当前块的模板划分为M个子模板。
在一些实施例中，模板确定单元22，具体用于将所述分界线向所述当前块的模板中进行延伸，得到所述分界线在所述当前块的模板中的延伸线；使用所述延伸线，将所述当前块的模板划分为M个矩形子模板。
在一些实施例中，模板确定单元22，具体用于将所述当前块的上方模板划分成P个子模板；和/或，将所述当前块的左侧模板划分成Q个子模板；其中，所述P和Q均为小于或等于M的整数，且所述P与Q之和等于M。
在一些实施例中，模板确定单元22，具体用于沿着竖直方向，将所述上方模板划分为P个子模板。
在一些实施例中，模板确定单元22，具体用于根据预设的像素点数，将所述上方模板划分成P个子模板。
在一些实施例中，模板确定单元22，具体用于将n列像素作为一个最小划分单元，将所述上方模板划分成P个子模板，所述n为正整数。
可选的，所述n根据所述当前块的长度确定。
在一些实施例中，模板确定单元22，具体用于沿着水平方向，将所述左侧模板划分为Q个子模板。
在一些实施例中，模板确定单元22，具体用于根据预设的像素点数，将所述左侧模板划分成Q个子模板。
在一些实施例中，模板确定单元22，具体用于将m行像素作为一个最小划分单元，将所述左侧模板划分成Q个 子模板，所述m为正整数。
可选的，所述m根据所述当前块的宽度确定。
在一些实施例中，模板确定单元22，具体用于对于所述M个子模板中的第j个子模板，根据所述权重导出模式，确定所述第j个子模板中的第一点关于第i个预测模式的权重，所述第i个预测模式为所述K个预测模式中的任一预测模式；根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中。
在一些实施例中，模板确定单元22，具体用于根据所述权重导出模式确定角度索引和距离索引；根据所述角度索引和距离索引，确定所述第j个子模板中的第一点关于第i个预测模式的权重。
在一些实施例中，所述第一点为所述第j个子模板与所述当前块的交界线上的一个点。
可选的，所述第一点为所述交界线的中点。
在一些实施例中，模板确定单元22，具体用于根据所述角度索引、所述距离索引和所述当前块的大小，确定所述第一点的第一参数，所述第一参数用于确定权重；根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重。
在一些实施例中，模板确定单元22，具体用于根据所述第一点的第一参数，确定所述第一点的第二参数；根据所述第一点的第二参数，确定所述第一点关于第i个预测模式的权重。
在一些实施例中，模板确定单元22，具体用于根据所述第一点的第一参数、第一预设值和第二预设值，确定所述第一点关于第i个预测模式的权重。
可选的，所述第一点关于第i个预测模式的权重为第一数值或为第二数值。
在一些实施例中，模板确定单元22，具体用于若所述第一点关于第i个预测模式的权重大于第一预设值，则将所述第j个子模板对应到第i个模板中，所述第i个模板为所述K个模板中的一个模板。
在一些实施例中，若所述K为2，i为1，模板确定单元22，具体用于若所述第一点关于第一预测模式的权重大于或等于第二预设值，则将所述第j个子模板对应到第一模板中；若所述第一点关于第一预测模式的权重小于第二预设值，则将所述第j个子模板对应到第二模板中。
在一些实施例中，模板确定单元22，具体用于从预设的不同块大小所对应的第一对应关系中，确定所述当前块对应的目标第一对应关系，所述第一对应关系包括不同角度索引或不同的权重导出模式与所述K个模板之间的对应关系；从所述目标第一对应关系中，确定所述权重导出模式对应的所述K个模板。
在一些实施例中，模式确定单元23，具体用于针对所述K个预测模式中的第i个预测模式，获取至少一个候选预测模式；使用所述候选预测模式对所述K个模板中的第i个模板进行预测，得到所述第i个模板的预测样本；根据所述第i个模板的预测值和重建值，确定所述候选预测模式的代价；根据所述至少一个候选预测模式的代价，确定所述第i个预测模式。
在一些实施例中，预测单元24，具体用于根据所述权重导出模式，确定权重；根据所述K个预测模式，确定K个预测值；根据所述权重对所述K个预测值进行加权，得到最终的预测值。
在一些实施例中，模式确定单元23，具体用于若所述K个模板所包括的点数均大于预设阈值时，则根据所述K个模板，确定K个预测模式。
在一些实施例中，模式确定单元23，还用于若所述K个模板中至少一个模板包括的点数小于预设阈值时，则根据所述权重导出模式，确定所述K个预测模式。
在一些实施例中，模式确定单元23，还用于将第一标志写入码流，所述第一标志用于指示是否采用模板匹配方式导出预测模式。
在一些实施例中，若所述K个预测模式中的至少一个为帧内预测模式时，则模式确定单元23，具体用于根据所述权重导出模式确定角度索引；将所述角度索引对应的帧内预测模式，确定为所述K个预测模式中的至少一个。
在一些实施例中，所述若所述K个预测模式中的至少一个为帧内预测模式时，模式确定单元23，具体用于获取所述权重导出模式对应的帧内预测模式；根据所述权重导出模式对应的帧内预测模式，确定所述K个预测模式中的至少一个。
在一些实施例中，所述权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式、与所述分界线垂直的帧内预测模式和planar模式中的至少一个。
在一些实施例中，若所述K个预测模式中的第i个预测模式为帧间预测模式时，则上述预测单元24，具体用于根据所述第i个预测模式，确定运动信息；根据所述运动信息，确定第i个预测值；根据所述K个预测模式中除所述第i个预测模式外的其他预测模式，确定K-1个预测值；根据所述权重导出模式确定权重；根据所述第i个预测值、所述K-1个预测值和所述权重，确定最终的预测值。
应理解，装置实施例与方法实施例可以相互对应，类似的描述可以参照方法实施例。为避免重复，此处不再赘述。具体地，图23所示的装置20可以对应于执行本申请实施例的编码端的预测方法中的相应主体，并且装置20中的各个单元的前述和其它操作和/或功能分别为了实现编码端的预测方法等各个方法中的相应流程，为了简洁，在此不再赘述。
上文中结合附图从功能单元的角度描述了本申请实施例的装置和系统。应理解，该功能单元可以通过硬件形式实现，也可以通过软件形式的指令实现，还可以通过硬件和软件单元组合实现。具体地，本申请实施例中的方法实施例的各步骤可以通过处理器中的硬件的集成逻辑电路和/或软件形式的指令完成，结合本申请实施例公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件单元组合执行完成。可选地，软件单元可以位于随机存储器，闪存、只读存储器、可编程只读存储器、电可擦写可编程存储器、寄存器等本领域的成熟的存储介质中。该存储介质位于存储器，处理器读取存储器中的信息，结合其硬件完成上述方法实施例中的步骤。
图24是本申请实施例提供的电子设备的示意性框图。
如图24所示，该电子设备30可以为本申请实施例所述的视频编码器，或者视频解码器，该电子设备30可包括：
存储器33和处理器32，该存储器33用于存储计算机程序34，并将该程序代码34传输给该处理器32。换言之， 该处理器32可以从存储器33中调用并运行计算机程序34，以实现本申请实施例中的方法。
例如，该处理器32可用于根据该计算机程序34中的指令执行上述方法200中的步骤。
在本申请的一些实施例中，该处理器32可以包括但不限于：
通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等等。
在本申请的一些实施例中，该存储器33包括但不限于：
易失性存储器和/或非易失性存储器。其中，非易失性存储器可以是只读存储器(Read-Only Memory，ROM)、可编程只读存储器(Programmable ROM，PROM)、可擦除可编程只读存储器(Erasable PROM，EPROM)、电可擦除可编程只读存储器(Electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(Random Access Memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用，例如静态随机存取存储器(Static RAM，SRAM)、动态随机存取存储器(Dynamic RAM，DRAM)、同步动态随机存取存储器(Synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synch link DRAM，SLDRAM)和直接内存总线随机存取存储器(Direct Rambus RAM，DR RAM)。
在本申请的一些实施例中，该计算机程序34可以被分割成一个或多个单元，该一个或者多个单元被存储在该存储器33中，并由该处理器32执行，以完成本申请提供的方法。该一个或多个单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述该计算机程序34在该电子设备30中的执行过程。
如图24所示，该电子设备30还可包括：
收发器33，该收发器33可连接至该处理器32或存储器33。
其中，处理器32可以控制该收发器33与其他设备进行通信，具体地，可以向其他设备发送信息或数据，或接收其他设备发送的信息或数据。收发器33可以包括发射机和接收机。收发器33还可以进一步包括天线，天线的数量可以为一个或多个。
应当理解，该电子设备30中的各个组件通过总线系统相连，其中，总线系统除包括数据总线之外，还包括电源总线、控制总线和状态信号总线。
图25是本申请实施例提供的视频编解码系统的示意性框图。
如图25所示，该视频编解码系统40可包括：视频编码器41和视频解码器42，其中视频编码器41用于执行本申请实施例涉及的视频编码方法，视频解码器42用于执行本申请实施例涉及的视频解码方法。
本申请还提供了一种计算机存储介质，其上存储有计算机程序，该计算机程序被计算机执行时使得该计算机能够执行上述方法实施例的方法。或者说，本申请实施例还提供一种包含指令的计算机程序产品，该指令被计算机执行时使得计算机执行上述方法实施例的方法。
本申请还提供了一种码流，该码流是根据上述编码方法生成的。
当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。该计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行该计算机程序指令时，全部或部分地产生按照本申请实施例该的流程或功能。该计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。该计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，该计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(digital subscriber line，DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。该计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。该可用介质可以是磁性介质(例如，软盘、硬盘、磁带)、光介质(例如数字视频光盘(digital video disc，DVD))、或者半导体介质(例如固态硬盘(solid state disk，SSD))等。
本领域普通技术人员可以意识到，结合本申请中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，该单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。例如，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
以上内容，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以该权利要求的保护范围为准。

Claims (77)

  1. 一种预测方法，应用于解码器，其特征在于，包括：
    解码码流，确定当前块的权重导出模式；
    根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
    根据所述K个模板，确定K个预测模式；
    根据所述K个预测模式和所述权重导出模式，确定预测值。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，包括：
    根据所述权重导出模式，将所述当前块的模板划分为所述K个模板。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述权重导出模式，将所述当前块的模板划分为所述K个模板，包括：
    将所述当前块的模板划分为M个子模板，所述M为大于或等于K的正整数；
    根据所述权重导出模式，将所述M个子模板对应到所述K个模板中。
  4. 根据权利要求3所述的方法，其特征在于，所述将所述当前块的模板划分为M个子模板，包括：
    根据所述权重导出模式，将所述当前块的模板划分为M个子模板。
  5. 根据权利要求4所述的方法，其特征在于，所述根据所述权重导出模式，将所述当前块的模板划分为M个子模板，包括：
    根据所述权重导出模式，确定权重的分界线；
    将所述分界线向所述当前块的模板中进行延长，以将所述当前块的模板划分为M个子模板。
  6. 根据权利要求5所述的方法，其特征在于，所述将所述分界线向所述当前块的模板中进行延长，以将所述当前块的模板划分为M个子模板，包括：
    将所述分界线向所述当前块的模板中进行延长，得到所述分界线在所述当前块的模板中的延伸线；
    使用所述延伸线，将所述当前块的模板划分为M个矩形子模板。
  7. 根据权利要求3所述的方法，其特征在于，所述将所述当前块的模板划分为M个子模板，包括：
    将所述当前块的上方模板划分成P个子模板；和/或，
    将所述当前块的左侧模板划分成Q个子模板；
    其中，所述P和Q均为小于或等于M的整数，且所述P与Q之和等于M。
  8. 根据权利要求7所述的方法，其特征在于，所述将所述当前块的上方模板划分成P个子模板，包括：
    沿着竖直方向，将所述上方模板划分为P个子模板。
  9. 根据权利要求7所述的方法，其特征在于，所述将所述当前块的上方模板划分成P个子模板，包括：
    根据预设的像素点数，将所述上方模板划分成P个子模板。
  10. 根据权利要求9所述的方法，其特征在于，所述根据预设的像素点数，将所述上方模板划分成P个子模板，包括：
    将n列像素作为一个最小划分单元，将所述上方模板划分成P个子模板，所述n为正整数。
  11. 根据权利要求10所述的方法，其特征在于，所述n根据所述当前块的长度确定。
  12. 根据权利要求7所述的方法，其特征在于，所述将所述当前块的左侧模板划分成Q个子模板，包括：
    沿着水平方向，将所述左侧模板划分为Q个子模板。
  13. 根据权利要求7所述的方法，其特征在于，所述将所述当前块的左侧模板划分成Q个子模板，包括：
    根据预设的像素点数，将所述左侧模板划分成Q个子模板。
  14. 根据权利要求13所述的方法，其特征在于，所述根据预设的像素点数，将所述左侧模板划分成Q个子模板，包括：
    将m行像素作为一个最小划分单元，将所述左侧模板划分成Q个子模板，所述m为正整数。
  15. 根据权利要求14所述的方法，其特征在于，所述m根据所述当前块的宽度确定。
  16. 根据权利要求3-15任一项所述的方法，其特征在于，根据所述权重导出模式，将所述M个子模板对应到所述K个模板中，包括：
    对于所述M个子模板中的第j个子模板，根据所述权重导出模式，确定所述第j个子模板中的第一点关于第i个预测模式的权重，所述第i个预测模式为所述K个预测模式中的任一预测模式；
    根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中。
  17. 根据权利要求16所述的方法，其特征在于，所述根据所述权重导出模式，确定所述第j个子模板中的第一点关于第i个预测模式的权重，包括：
    根据所述权重导出模式确定角度索引和距离索引；
    根据所述角度索引和距离索引，确定所述第j个子模板中的第一点关于第i个预测模式的权重。
  18. 根据权利要求16所述的方法，其特征在于，所述第一点为所述第j个子模板与所述当前块的交界线上的一个点。
  19. 根据权利要求18所述的方法，其特征在于，所述第一点为所述交界线的中点。
  20. 根据权利要求17所述的方法，其特征在于，所述根据所述角度索引和距离索引，确定所述第j个子模板中的第一点关于第i个预测模式的权重，包括：
    根据所述角度索引、所述距离索引和所述当前块的大小，确定所述第一点的第一参数，所述第一参数用于确定权重；
    根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重。
  21. 根据权利要求20所述的方法，其特征在于，所述根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重，包括：
    根据所述第一点的第一参数，确定所述第一点的第二参数，所述第二参数用于确定权重；
    根据所述第一点的第二参数，确定所述第一点关于第i个预测模式的权重。
  22. 根据权利要求20所述的方法，其特征在于，所述根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重，包括：
    根据所述第一点的第一参数、第一预设值和第二预设值，确定所述第一点关于第i个预测模式的权重。
  23. 根据权利要求22所述的方法，其特征在于，所述第一点关于第i个预测模式的权重为第一数值或为第二数值。
  24. 根据权利要求16所述的方法，其特征在于，所述根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中，包括：
    若所述第一点关于第i个预测模式的权重大于第一预设值，则将所述第j个子模板对应到第i个模板中，所述第i个模板为所述K个模板中的一个模板。
  25. 根据权利要求24所述的方法，其特征在于，若所述K为2，i为1，所述根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中，包括：
    若所述第一点关于第一预测模式的权重大于或等于第二预设值，则将所述第j个子模板对应到第一模板中；
    若所述第一点关于第一预测模式的权重小于第二预设值，则将所述第j个子模板对应到第二模板中。
  26. 根据权利要求1所述的方法，其特征在于，所述根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，包括：
    从预设的不同块大小所对应的第一对应关系中，确定所述当前块对应的目标第一对应关系，所述第一对应关系包括不同角度索引或不同的权重导出模式与所述K个模板之间的对应关系；
    从所述目标第一对应关系中，确定所述权重导出模式对应的所述K个模板。
  27. 根据权利要求1-15、17-26任一项所述的方法，其特征在于，所述根据所述K个模板，确定K个预测模式，包括：
    针对所述K个预测模式中的第i个预测模式，获取至少一个候选预测模式；
    使用所述候选预测模式对所述K个模板中的第i个模板进行预测，得到所述第i个模板的预测值；
    根据所述第i个模板的预测值和重建值，确定所述候选预测模式的代价；
    根据所述至少一个候选预测模式的代价，确定所述第i个预测模式。
  28. 根据权利要求1-15、17-26任一项所述的方法，其特征在于，所述根据所述K个预测模式和所述权重导出模式，确定预测值，包括：
    根据所述权重导出模式，确定权重；
    根据所述K个预测模式，确定K个预测值；
    根据所述权重对所述K个预测值进行加权，得到所述预测值。
  29. 根据权利要求1-15、17-26任一项所述的方法，其特征在于，所述根据所述K个模板，确定K个预测模式，包括：
    若所述K个模板所包括的点数均大于预设阈值时，则根据所述K个模板，确定K个预测模式。
  30. 根据权利要求29所述的方法，其特征在于，所述方法还包括：
    若所述K个模板中至少一个模板包括的点数小于预设阈值时，则根据所述权重导出模式，确定所述K个预测模式。
  31. 根据权利要求1-15、17-26任一项所述的方法，其特征在于，所述方法还包括：
    解码所述码流，得到第一标志，所述第一标志用于指示是否采用模板匹配方式导出预测模式；
    所述根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，包括：
    若所述第一标志指示采用模板匹配方式导出预测模式时，则根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板。
  32. 根据权利要求31所述的方法，其特征在于，所述方法还包括：
    若所述第一标志指示不采用模板匹配方式导出预测模式时，则根据所述权重导出模式，确定所述K个预测模式。
  33. 根据权利要求30或32所述的方法，其特征在于，所述若所述K个预测模式中的至少一个为帧内预测模式时，则根据所述权重导出模式，确定所述K个预测模式，包括：
    根据所述权重导出模式确定角度索引；
    将所述角度索引对应的帧内预测模式，确定为所述K个预测模式中的至少一个。
  34. 根据权利要求30或32所述的方法，其特征在于，所述若所述K个预测模式中的至少一个为帧内预测模式时，根据所述权重导出模式，确定所述K个预测模式，包括：
    获取所述权重导出模式对应的帧内预测模式；
    根据所述权重导出模式对应的帧内预测模式，确定所述K个预测模式中的至少一个。
  35. 根据权利要求34所述的方法，其特征在于，所述权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式、与所述分界线垂直的帧内预测模式和planar模式中的至少一个。
  36. 根据权利要求1-15、17-26任一项所述的方法，其特征在于，若所述K个预测模式中的第i个预测模式为帧间预测模式时，所述根据所述K个预测模式和所述权重导出模式，确定预测值，包括：
    根据所述第i个预测模式，确定运动信息；
    根据所述运动信息，确定第i个预测值；
    根据所述K个预测模式中除所述第i个预测模式外的其他预测模式，确定K-1个预测值；
    根据所述权重导出模式确定权重；
    根据所述第i个预测值、所述K-1个预测值和所述权重，确定所述预测值。
  37. 一种预测方法，应用于编码器，其特征在于，包括：
    确定当前块的权重导出模式；
    根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
    根据所述K个模板，确定K个预测模式；
    根据所述K个预测模式和所述权重导出模式，确定预测值。
  38. 根据权利要求37所述的方法，其特征在于，所述根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，包括：
    根据所述权重导出模式，将所述当前块的模板划分为所述K个模板。
  39. 根据权利要求38所述的方法，其特征在于，所述根据所述权重导出模式，将所述当前块的模板划分为所述K个模板，包括：
    将所述当前块的模板划分为M个子模板，所述M为大于或等于K的正整数；
    根据所述权重导出模式，将所述M个子模板对应到所述K个模板中。
  40. 根据权利要求39所述的方法，其特征在于，所述将所述当前块的模板划分为M个子模板，包括：
    根据所述权重导出模式，将所述当前块的模板划分为M个子模板。
  41. 根据权利要求40所述的方法，其特征在于，所述根据所述权重导出模式，将所述当前块的模板划分为M个子模板，包括：
    根据所述权重导出模式，确定权重的分界线；
    将所述分界线向所述当前块的模板中进行延长，以将所述当前块的模板划分为M个子模板。
  42. 根据权利要求41所述的方法，其特征在于，所述将所述分界线向所述当前块的模板中进行延长，以将所述当前块的模板划分为M个子模板，包括：
    将所述分界线向所述当前块的模板中进行延长，得到所述分界线在所述当前块的模板中的延伸线；
    使用所述延伸线，将所述当前块的模板划分为M个矩形子模板。
  43. 根据权利要求39所述的方法，其特征在于，所述将所述当前块的模板划分为M个子模板，包括：
    将所述当前块的上方模板划分成P个子模板；和/或，
    将所述当前块的左侧模板划分成Q个子模板；
    其中，所述P和Q均为小于或等于M的整数，且所述P与Q之和等于M。
  44. 根据权利要求43所述的方法，其特征在于，所述将所述当前块的上方模板划分成P个子模板，包括：
    沿着竖直方向，将所述上方模板划分为P个子模板。
  45. 根据权利要求43所述的方法，其特征在于，所述将所述当前块的上方模板划分成P个子模板，包括：
    根据预设的像素点数，将所述上方模板划分成P个子模板。
  46. 根据权利要求45所述的方法，其特征在于，所述根据预设的像素点数，将所述上方模板划分成P个子模板，包括：
    将n列像素作为一个最小划分单元，将所述上方模板划分成P个子模板，所述n为正整数。
  47. 根据权利要求46所述的方法，其特征在于，所述n根据所述当前块的长度确定。
  48. 根据权利要求43所述的方法，其特征在于，所述将所述当前块的左侧模板划分成Q个子模板，包括：
    沿着水平方向，将所述左侧模板划分为Q个子模板。
  49. 根据权利要求43所述的方法，其特征在于，所述将所述当前块的左侧模板划分成Q个子模板，包括：
    根据预设的像素点数，将所述左侧模板划分成Q个子模板。
  50. 根据权利要求49所述的方法，其特征在于，所述根据预设的像素点数，将所述左侧模板划分成Q个子模板，包括：
    将m行像素作为一个最小划分单元，将所述左侧模板划分成Q个子模板，所述m为正整数。
  51. 根据权利要求50所述的方法，其特征在于，所述m根据所述当前块的宽度确定。
  52. 根据权利要求39-51任一项所述的方法，其特征在于，根据所述权重导出模式，将所述M个子模板对应到所述K个模板中，包括：
    对于所述M个子模板中的第j个子模板，根据所述权重导出模式，确定所述第j个子模板中的第一点关于第i个预测模式的权重，所述第i个预测模式为所述K个预测模式中的任一预测模式；
    根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中。
  53. 根据权利要求52所述的方法，其特征在于，所述根据所述权重导出模式，确定所述第j个子模板中的第一点关于第i个预测模式的权重，包括：
    根据所述权重导出模式确定角度索引和距离索引；
    根据所述角度索引和距离索引，确定所述第j个子模板中的第一点关于第i个预测模式的权重。
  54. 根据权利要求52所述的方法，其特征在于，所述第一点为所述第j个子模板与所述当前块的交界线上的一个点。
  55. 根据权利要求54所述的方法，其特征在于，所述第一点为所述交界线的中点。
  56. 根据权利要求53所述的方法，其特征在于，所述根据所述角度索引和距离索引，确定所述第j个子模板中的第一点关于第i个预测模式的权重，包括：
    根据所述角度索引、所述距离索引和所述当前块的大小，确定所述第一点的第一参数，所述第一参数用于确定权重；
    根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重。
  57. 根据权利要求56所述的方法，其特征在于，所述根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重，包括：
    根据所述第一点的第一参数，确定所述第一点的第二参数，所述第二参数用于确定权重；
    根据所述第一点的第二参数，确定所述第一点关于第i个预测模式的权重。
  58. 根据权利要求57所述的方法，其特征在于，所述根据所述第一点的第一参数，确定所述第一点关于第i个预测模式的权重，包括：
    根据所述第一点的第一参数、第一预设值和第二预设值，确定所述第一点关于第i个预测模式的权重。
  59. 根据权利要求58所述的方法，其特征在于，所述第一点关于第i个预测模式的权重为第一数值或为第二数值。
  60. 根据权利要求52所述的方法，其特征在于，所述根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中，包括：
    若所述第一点关于第i个预测模式的权重大于第一预设值，则将所述第j个子模板对应到第i个模板中，所述第i个模板为所述K个模板中的一个模板。
  61. 根据权利要求60所述的方法，其特征在于，若所述K为2，i为1，所述根据所述第j个子模板中的第一点关于第i个预测模式的权重，将所述第j个子模板对应到所述K个模板中，包括：
    若所述第一点关于第一预测模式的权重大于或等于第二预设值，则将所述第j个子模板对应到第一模板中；
    若所述第一点关于第一预测模式的权重小于第二预设值，则将所述第j个子模板对应到第二模板中。
  62. 根据权利要求37所述的方法，其特征在于，所述根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，包括：
    从预设的不同块大小所对应的第一对应关系中，确定所述当前块对应的目标第一对应关系，所述第一对应关系包括不同角度索引或不同的权重导出模式与所述K个模板之间的对应关系；
    从所述目标第一对应关系中，确定所述权重导出模式对应的所述K个模板。
  63. 根据权利要求37-51、53-62任一项所述的方法，其特征在于，所述根据所述K个模板，确定K个预测模式，包括：
    针对所述K个预测模式中的第i个预测模式，获取至少一个候选预测模式；
    使用所述候选预测模式对所述K个模板中的第i个模板进行预测，得到所述第i个模板的预测样本；
    根据所述第i个模板的预测值和重建值，确定所述候选预测模式的代价；
    根据所述至少一个候选预测模式的代价，确定所述第i个预测模式。
  64. 根据权利要求37-51、53-62任一项所述的方法，其特征在于，所述根据所述K个预测模式和所述权重导出模式，确定预测值，包括：
    根据所述权重导出模式，确定权重；
    根据所述K个预测模式，确定K个预测值；
    根据所述权重对所述K个预测值进行加权，得到最终的预测值。
  65. 根据权利要求37-51、53-62任一项所述的方法，其特征在于，所述根据所述K个模板，确定K个预测模式，包括：
    若所述K个模板所包括的点数均大于预设阈值时，则根据所述K个模板，确定K个预测模式。
  66. 根据权利要求65所述的方法，其特征在于，所述方法还包括：
    若所述K个模板中至少一个模板包括的点数小于预设阈值时，则根据所述权重导出模式，确定所述K个预测模式。
  67. 根据权利要求65所述的方法，其特征在于，所述方法还包括：
    将第一标志写入码流，所述第一标志用于指示是否采用模板匹配方式导出预测模式。
  68. 根据权利要求66所述的方法，其特征在于，所述若所述K个预测模式中的至少一个为帧内预测模式时，则根据所述权重导出模式，确定所述K个预测模式，包括：
    根据所述权重导出模式确定角度索引；
    将所述角度索引对应的帧内预测模式，确定为所述K个预测模式中的至少一个。
  69. 根据权利要求66或68所述的方法，其特征在于，所述若所述K个预测模式中的至少一个为帧内预测模式时，根据所述权重导出模式，确定所述K个预测模式，包括：
    获取所述权重导出模式对应的帧内预测模式；
    根据所述权重导出模式对应的帧内预测模式，确定所述K个预测模式中的至少一个。
  70. 根据权利要求69所述的方法，其特征在于，所述权重导出模式对应的帧内预测模式包括与权重的分界线平行的帧内预测模式、与所述分界线垂直的帧内预测模式和planar模式中的至少一个。
  71. 根据权利要求37-51、53-62任一项所述的方法，其特征在于，若所述K个预测模式中的第i个预测模式为帧间预测模式时，所述根据所述K个预测模式和所述权重导出模式，确定预测值，包括：
    根据所述第i个预测模式，确定运动信息；
    根据所述运动信息，确定第i个预测值；
    根据所述K个预测模式中除所述第i个预测模式外的其他预测模式，确定K-1个预测值；
    根据所述权重导出模式确定权重；
    根据所述第i个预测值、所述K-1个预测值和所述权重，确定最终的预测值。
  72. 一种预测装置，应用于解码器，其特征在于，包括：
    解码单元，用于解码码流，确定当前块的权重导出模式；
    模板确定单元，用于根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
    模式确定单元，用于根据所述K个模板，确定K个预测模式；
    预测单元，用于根据所述K个预测模式和所述权重导出模式，确定预测值。
  73. 一种预测装置，应用于编码器，其特征在于，包括：
    确定单元，用于确定当前块的权重导出模式；
    模板确定单元，用于根据所述当前块的大小和所述权重导出模式中的至少一个，确定K个模板，所述K为大于1的正整数；
    模式确定单元，用于根据所述K个模板，确定K个预测模式；
    预测单元，用于根据所述K个预测模式和所述权重导出模式，确定预测值。
  74. 一种电子设备，其特征在于，包括处理器和存储器；
    所示存储器用于存储计算机程序；
    所述处理器用于调用并运行所述存储器中存储的计算机程序，以实现上述权利要求1至36或37至71任一项所述的方法。
  75. 一种视频编解码系统，其特征在于，包括：视频编码器和视频解码器；
    所述的视频解码器用于实现上述权利要求1至36任一项所述的方法；
    所述的视频编码器用于实现上述权利要求37至71任一项所述的方法。
  76. 一种计算机可读存储介质，其特征在于，用于存储计算机程序；
    所述计算机程序使得计算机执行如上述权利要求1至36或37至71任一项所述的方法。
  77. 一种码流，其特征在于，所述码流是基于如上述权利要求37至71任一项所述的方法生成的。
PCT/CN2021/143977 2021-12-31 2021-12-31 预测方法、装置、设备、系统、及存储介质 WO2023123478A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2021/143977 WO2023123478A1 (zh) 2021-12-31 2021-12-31 预测方法、装置、设备、系统、及存储介质
CN202180105280.0A CN118476224A (zh) 2021-12-31 2021-12-31 预测方法、装置、设备、系统、及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/143977 WO2023123478A1 (zh) 2021-12-31 2021-12-31 预测方法、装置、设备、系统、及存储介质

Publications (1)

Publication Number Publication Date
WO2023123478A1 true WO2023123478A1 (zh) 2023-07-06

Family

ID=86997169

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/143977 WO2023123478A1 (zh) 2021-12-31 2021-12-31 预测方法、装置、设备、系统、及存储介质

Country Status (2)

Country Link
CN (1) CN118476224A (zh)
WO (1) WO2023123478A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020180159A1 (ko) * 2019-03-06 2020-09-10 엘지전자 주식회사 영상 부호화/복호화 방법, 장치 및 비트스트림을 전송하는 방법
CN113709498A (zh) * 2020-05-20 2021-11-26 Oppo广东移动通信有限公司 帧间预测方法、编码器、解码器以及计算机存储介质
CN113709500A (zh) * 2019-12-23 2021-11-26 杭州海康威视数字技术股份有限公司 一种编解码方法、装置及其设备
WO2021238396A1 (zh) * 2020-05-29 2021-12-02 Oppo广东移动通信有限公司 帧间预测方法、编码器、解码器以及计算机存储介质
CN113840148A (zh) * 2020-06-24 2021-12-24 Oppo广东移动通信有限公司 帧间预测方法、编码器、解码器以及计算机存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020180159A1 (ko) * 2019-03-06 2020-09-10 엘지전자 주식회사 영상 부호화/복호화 방법, 장치 및 비트스트림을 전송하는 방법
CN113709500A (zh) * 2019-12-23 2021-11-26 杭州海康威视数字技术股份有限公司 一种编解码方法、装置及其设备
CN113709498A (zh) * 2020-05-20 2021-11-26 Oppo广东移动通信有限公司 帧间预测方法、编码器、解码器以及计算机存储介质
WO2021238396A1 (zh) * 2020-05-29 2021-12-02 Oppo广东移动通信有限公司 帧间预测方法、编码器、解码器以及计算机存储介质
CN113840148A (zh) * 2020-06-24 2021-12-24 Oppo广东移动通信有限公司 帧间预测方法、编码器、解码器以及计算机存储介质

Also Published As

Publication number Publication date
CN118476224A (zh) 2024-08-09

Similar Documents

Publication Publication Date Title
CN110999291B (zh) 用于划分视频数据的帧间预测片段中的视频块的系统和方法
CN112204967B (zh) 视频数据编码的设备和方法
AU2018282523A1 (en) Intra filtering applied together with transform processing in video coding
TW202127883A (zh) 視訊編解碼中具有經簡化運動場儲存及運動補償的幾何分割模式
CN112789858B (zh) 帧内预测方法及设备
CN115695783A (zh) 对图像的块进行解码的方法和设备，解码设备，和计算机可读介质
US20240187624A1 (en) Methods and devices for decoder-side intra mode derivation
US20230319267A1 (en) Video coding method and video decoder
WO2023044868A1 (zh) 视频编解码方法、设备、系统、及存储介质
WO2023123478A1 (zh) 预测方法、装置、设备、系统、及存储介质
WO2022271756A1 (en) Video coding using multi-direction intra prediction
WO2023123495A1 (zh) 预测方法、装置、设备、系统、及存储介质
EP4324208A1 (en) Video coding using multi-model linear model
WO2023197433A1 (zh) 视频编解码方法、装置、设备、系统、及存储介质
WO2024007128A1 (zh) 视频编解码方法、装置、设备、系统、及存储介质
WO2024077553A1 (zh) 视频编解码方法、装置、设备、系统、及存储介质
WO2024108391A1 (zh) 视频编解码方法、装置、设备、系统、及存储介质
WO2024152254A1 (zh) 视频编解码方法、装置、设备、系统、及存储介质
WO2023122968A1 (zh) 帧内预测方法、设备、系统、及存储介质
WO2024183007A1 (zh) 视频编解码方法、装置、设备、系统、及存储介质
WO2023197229A1 (zh) 视频编解码方法、装置、设备、系统及存储介质
WO2024192733A1 (zh) 视频编解码方法、装置、设备、系统、及存储介质
WO2023122969A1 (zh) 帧内预测方法、设备、系统、及存储介质
WO2023220970A1 (zh) 视频编码方法、装置、设备、系统、及存储介质
TW202433936A (zh) 視訊編解碼方法、裝置、設備、系統、及儲存媒介

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: 21969823

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE