RU2808608C2 - Early termination of optical flow refinement - Google Patents

Abstract

FIELD: image processing.

SUBSTANCE: invention is related to refinement of optical flow. A video encoding method is claimed, implemented in a decoding device or encoding device, wherein the method comprises: obtaining initial motion vectors for the current block; obtaining first predictions for the sample value in the current block based on the original motion vectors; calculating a first matching value according to the first predictions; determining whether the optical flow refinement process should be performed or not according to at least one preset condition, wherein the at least one preset condition comprises a condition of whether the calculated first matching value is equal to or greater than a threshold value; and performing an optical flow refinement process to obtain a final inter-prediction for the sample value in the current block when it is determined that the optical flow refinement process is to be performed.

EFFECT: increase of accuracy of external prediction of the current block of the picture to be encoded.

17 cl, 11 dwg

Description

Field of technology

The present invention relates to the field of image processing and, in particular, to optical flow refinement.

State of the art

Video encoding (video encoding and decoding) is used in a wide range of digital video applications, such as digital broadcast television, video transmission over the Internet and mobile networks, and real-time conversational applications such as video chat, video conferencing, DVD and Blu-ray discs. , systems for collecting and editing video content and video cameras for security applications.

The amount of video data required to illustrate even a relatively short video can be significant, which can lead to difficulties when the data must be streamed or otherwise transmitted over a communications network with limited bandwidth. Thus, video data is typically compressed before being transmitted over modern telecommunications networks. Video size can also be an issue when the video is stored on a storage device, as memory resources may be limited. Video compression devices often use software and/or hardware at the source to encode video data before transmission or storage, thereby reducing the amount of data required to represent digital video images. The compressed data is then received at its destination by a video decompression device, which decodes the video data. With limited network resources and ever-increasing demands for higher video quality, improved compression and decompression techniques are desired that improve compression ratios with little or no loss in image quality.

Recently, inter-prediction coding has been improved by refining the optical flow with bidirectional prediction. This method can improve the accuracy of the external prediction of the current block of the picture to be encoded. However, refining (improving) optical flow with bidirectional prediction is relatively expensive in terms of computational load. Thus, a trade-off must be made between accurate external prediction and computational load. The present invention solves this problem.

The essence of the invention

Embodiments of the present application provide apparatus and methods for encoding and decoding according to independent claims.

The above and other objects are achieved by the essence of the independent claims. Additional forms of implementation are obvious from the dependent claims, description and figures.

A video encoding method implemented in a decoding device or an encoding device is provided, comprising:

obtaining initial motion vectors for the current block (eg, an encoding block, or a prediction block, or a subblock);

obtaining first predictions (two prediction values for the outer prediction) for the sample value in the current block based on the original motion vectors;

calculating a first matching cost (eg, consisting of or containing some measure of similarity (or dissimilarity; see also detailed description below) according to the first predictions;

determining whether the optical flow refinement process should be performed or not according to at least one predetermined condition, wherein the at least one predetermined condition comprises a condition whether the calculated first matching cost is equal to or greater than a threshold value;

performing an optical flow refinement process to obtain a final inter-prediction for the sample value in the current block when it is determined that the optical flow refinement process is to be performed.

Thus, according to the invention, performing optical flow refinement, in particular bidirectional optical flow refinement, is performed on a conditional basis. Relatively expensive optical flow refinement is performed only under certain circumstances that provide a suitable desired improvement in the accuracy of the overall inter-prediction process. If it is determined that the optical flow refinement is not likely to result in an increase in inter-prediction accuracy that justifies the relatively high computational load required to perform the optical flow refinement, the optical flow refinement may be suppressed. In this way, decoding time can be significantly reduced. The original motion vectors can be signaled in the bit stream. Alternatively, motion vector predictions and motion vector difference components may be provided for the original motion vectors.

For example, the at least one predefined condition comprises a condition that the current block can be predicted by motion vector refinement at the decoder side. This particular condition, contained in at least one predefined condition, can be checked first to avoid unnecessary computational effort.

According to a particular embodiment, it is determined that the optical flow refinement process is to be performed when it is determined that all of at least one predetermined condition are satisfied. The at least one predefined condition may in principle contain one or more additional conditions. For example, the at least one predefined condition may include a condition that a particular flag is set (eg, to 1) to perform the optical refinement process. If not all conditions are met, no optical flow refinement can be performed at all according to a particular embodiment in order to reduce computational requirements.

The first predictions for the sample value in the current block can be obtained based on the first interpolation filter to achieve some sub-pixel precision. In particular, the first interpolation filter may be a relatively simple bilinear interpolation filter that provides fast filter processing.

The inventive video encoding method, implemented in a decoding device or encoding device, may contain some motion vector refinement other than optical flow refinement. Thus, the method may comprise obtaining refined motion vectors based on the original motion vectors and the cost of the first mapping; obtaining second predictions for the sample value in the current block according to the refined motion vectors when it is determined that an optical flow refinement process is to be performed, performing optical flow refinement based on the second predictions (representing the already refined predictions). The overall accuracy of the external prediction process can be improved by using refined motion vectors.

It should be noted that the first predictions as well as the first matching cost have already been calculated to refine the motion vector. Therefore, no additional computation is required to decide whether to stop/suppress the optical flow refinement process early, but the results of previous computations involved in the motion vector refinement process can be reused.

In each of the above-described embodiments of the method according to the invention, obtaining first predictions for the sample value in the current block based on the original motion vectors may comprise obtaining a number of candidate pairs based on the original motion vectors and obtaining first predictions for the sample value in the current block based on at least one of the candidate pairs, and calculating a first matching cost according to the first predictions may comprise determining a matching cost for each of the candidate pairs based on the first predictions and determining the smallest matching cost of the determined matching costs as the first matching cost.

During motion vector refinement, a number of candidate pairs for the refined motion vectors may be obtained, the pairs including a pair of original motion vectors. For example, candidate pairs for refined motion vectors contain a pair of original motion vectors (MV0, MV1) and pairs (MV0 + (0,1), MV1 + (0, -1)), (MV0 + (1,0), MV1 + (-1.0)), (MV0 + (0, -1), MV1 + (0.1)), (MV0 + (-1.0), MV1 + (1.0 )), where (1, -1) denotes a vector that has an offset of 1 in the horizontal (or x) direction and an offset of -1 in the vertical (or y) direction. For each of the pairs, a matching cost corresponding to that pair may be determined, and the aforementioned first matching cost may be determined to be the lesser of the matching costs determined for the candidate pairs for the refined motion vectors. According to specific examples, this could be the matching cost corresponding to a pair of source motion vectors (MV0, MV1) or (MV0, MV1) or (MV0'=MV0 + (0,1), MV1'=MV1 + (0,-1) ) with refined motion vectors MV0' and MV1'.

Using this kind of first match cost can be beneficial from an overall encoding perspective.

The above second predictions for the sample value in the current block may be obtained according to the second interpolation filter. This second interpolation filter may be a 6- or 8-tap interpolation filter, which is relatively expensive but beneficial in terms of subpixel accuracy.

The above refined motion vectors may be obtained according to a second matching cost to control the suitability of the refined motion vectors for inter-prediction. When the value of the second matching cost is greater than or equal to another threshold value, it may be determined that the optical flow refinement process should be performed. Otherwise, it may be determined that it is not worth performing any optical flow refinement processing.

According to another embodiment, only when it is determined that the optical flow refinement process should not be performed, the final external prediction is obtained by a weighted sum of the second predictions. The weighted sum of the second predictions provides some precision that may be considered sufficient in cases where it is not considered appropriate to perform the relatively expensive optical flow refinement process.

In general, the threshold or other threshold value may be a value that is calculated based on the bit depth of the first predictions. In addition, the threshold value can be obtained according to the number of predicted samples, which are used to calculate the first matching cost according to the first predictions. Additionally, the threshold value can be obtained according to the size (width and height in pixel units) of the current block. For example, the threshold could be th=nCbW x nCbH x K, where K is a value greater than zero, nCbW and nCbH are the width and height of the current block. For example, K=2.

In addition, the above-mentioned second matching cost may be an inferred cost obtained using matching costs estimated during motion vector refinement and a predetermined model for the matching cost shape near the position of the minimum matching cost. The predefined model in this context may be a linear combination model. Using a predefined model for the matching cost shape near the position of the minimum matching cost can improve the accuracy of the external prediction process.

The method according to all of the above-described embodiments may further comprise the step of generating an inter prediction block containing a final inter prediction for the sample value in the current block.

In addition, an encoder or decoder is provided containing some processing circuitry for performing the method according to any of the above-described embodiments. Also provided is a computer program product containing program code for performing a method according to any of the embodiments described above.

All of the video encoding method options described above can be implemented in a decoder or encoder. Thus, a decoder or encoder is provided containing:

one or more processors and a non-transitory computer-readable storage medium coupled to the processors and storing a program for execution by the processors, wherein the program, when executed by the processors, configures a decoder to perform a method in accordance with any of the above-described embodiments.

All of the above-described embodiments of the video encoding method can be implemented in an apparatus for use in an image encoder and/or an image decoder to satisfy the above need. Thus, an apparatus is provided for use in an image encoder and/or an image decoder, the apparatus comprising a source motion vector block configured to obtain source motion vectors for a current block (eg, an encoding block or a prediction block or a subblock); a first prediction block configured to obtain first predictions for the sample value in the current block based on the original motion vectors; a first matching cost calculating unit, configured to calculate a first matching cost (eg, a measure of similarity or dissimilarity) according to the first predictions; an optical flow refinement process determining unit, configured to determine whether the optical flow refinement process should be performed according to at least one predetermined condition, wherein the at least one predetermined condition comprises a condition whether the calculated first matching cost is equal to or greater than a certain threshold ; and an optical flow refinement process execution unit configured to perform an optical flow refinement process to obtain a final inter-prediction for the sample value in the current block when it is determined that the optical flow refinement process is to be executed.

This device, as described above, and with its variations described below, provides the same advantages as the methods described above.

The at least one predefined condition may include a condition that the current block can be predicted by motion vector refinement at the decoder side.

The optical flow refinement process determining unit may be configured to determine that the optical flow refinement process should be executed when it is determined that all of at least one predetermined condition are satisfied.

The apparatus may include a first interpolation filter (eg, a bilinear interpolation filter), and the first prediction block may be configured to obtain first predictions for the sample value in the current block by the first interpolation filter.

The device may further comprise a refined motion vector unit configured to obtain refined motion vectors based on the original motion vectors and the cost of the first mapping;

a second prediction unit configured to obtain second predictions for the sample value in the current block according to the refined motion vectors; And

the optical flow refinement process executing unit may be configured to perform optical flow refinement based on the second predictions when the optical flow refinement process determining unit determines that the optical flow refinement process should be executed.

In the above-described embodiments of the device, the first prediction block may be configured to obtain first predictions for the sample value in the current block based on the original motion vectors by obtaining a number of candidate pairs based on the original motion vectors and obtaining first predictions for the sample value in the current block based on at least one of the candidate pairs. In addition, the first matching cost calculating unit may be configured to calculate a first matching cost according to the first predictions by determining a matching cost for each of the candidate pairs based on the first predictions and determining the smallest matching cost of the determined matching costs as the first matching cost.

According to an embodiment, the device may further comprise a second interpolation filter (e.g., a relatively expensive 6- or 8-tap interpolation filter with relatively high subpixel precision), and the second prediction block may be configured to make second predictions for the sample value in the current block using the second interpolation filter.

According to another embodiment, the apparatus further comprises a second matching cost calculation unit configured to calculate the second matching cost, and wherein the refined motion vector unit is configured to obtain refined motion vectors according to the second matching cost. In this case, the optical flow refinement process determining unit may be configured to determine that the optical flow refinement process should be executed when the value of the second matching cost is greater than or equal to another threshold value.

The apparatus may further comprise a weighted sum predictor configured to obtain a final inter-prediction by the weighted sum of the second predictions only when it is determined by the optical flow refinement process determination unit that the optical flow refinement process should not be performed.

Additionally, the device may include a threshold calculation unit configured to calculate a threshold value or other threshold based on the bit depth of the first predictions. In addition, the device may further comprise a threshold calculating unit configured to calculate a threshold according to the number of predicted samples that are used to calculate the first matching cost according to the first predictions by the first matching cost computing unit. In addition, the device may further comprise a threshold calculation block configured to calculate a threshold according to the size of the current block. For example, the threshold could be th=nCbW x nCbH x K, where K is a value greater than zero, nCbW and nCbH are the width and height of the current block. For example, K=2.

According to specific embodiments, the second matching cost calculation block is configured to calculate the second matching cost as an inferred cost obtained using matching costs estimated during motion vector refinement performed by the motion vector refinement block and a predefined model (e.g., a linear combinational model) for the shape matching cost around the minimum matching cost position.

The apparatus according to any of the above-described embodiments may further comprise an inter prediction block generation unit configured to generate an inter prediction block containing a final inter prediction for the sample value in the current block.

Details of one or more embodiments are set forth in the accompanying drawings and description below. Other features, objects and advantages will be apparent from the description, drawings and claims.

BRIEF DESCRIPTION OF THE DRAWINGS

In the following embodiments, the inventions are described in more detail with reference to the accompanying figures and drawings, in which:

Fig. 1A is a block diagram showing an example of a video encoding system configured to implement embodiments of the invention;

FIG. 1B is a block diagram showing another example of a video encoding system configured to implement embodiments of the invention; FIG.

Fig. 2 is a block diagram showing an example of a video encoder configured to implement embodiments of the invention;

Fig. 3 is a block diagram showing an exemplary structure of a video decoder configured to implement embodiments of the invention;

Fig. 4 is a block diagram illustrating an example of encoding equipment or decoding equipment;

Fig. 5 is a block diagram illustrating another example of encoding equipment or decoding equipment;

FIG. 6 is a block diagram illustrating an embodiment of an optical refinement process; FIG.

FIG. 7 is a block diagram illustrating another embodiment of the optical refinement process; FIG.

FIG. 8 is a block diagram illustrating another embodiment of the optical refinement process; FIG.

FIG. 9 is a block diagram illustrating another embodiment of the optical refinement process.

Fig. 10 is a block diagram illustrating a video encoding method implemented in a decoding apparatus or an encoding apparatus according to an embodiment of the present invention.

Fig. 11 illustrates an apparatus for use in an image encoder and/or an image decoder according to an embodiment of the present invention.

The following identical reference numerals refer to identical or at least functionally equivalent features unless expressly indicated otherwise.

DETAILED DESCRIPTION OF IMPLEMENTATION OPTIONS

In the following description, reference is made to the accompanying drawings, which form a part of the disclosure and show by way of illustration specific aspects of embodiments of the invention or specific aspects in which embodiments of the present invention may be used. It is understood that embodiments of the invention may be used in other aspects and contain structural or logical changes not depicted in the figures. Therefore, the following detailed description should not be taken in a limiting sense, and the scope of the present invention is defined by the appended claims.

For example, it will be understood that the disclosure associated with the described method may also apply to a corresponding device or system configured to perform the method, and vice versa. For example, if one or a plurality of specific method steps are described, the corresponding apparatus may include one or a plurality of blocks, such as functional blocks, to perform the described one or multiple method steps (e.g., one block performing one or a plurality of steps, or a plurality of blocks , each of which performs one or more of a plurality of steps), even if such one or more blocks are not explicitly described or illustrated in the figures. On the other hand, for example, if a particular device is described in terms of one or a plurality of blocks, such as functional blocks, the corresponding method may include one step for performing the functionality of the one or multiple blocks (e.g., one step performing the functionality of the one or multiple blocks , or a plurality of steps, each of which performs the functionality of one or more of a plurality of blocks), even if such one or multiple steps are not explicitly described or illustrated in the figures. It is further understood that features of various exemplary embodiments and/or aspects described herein may be combined with each other unless specifically stated otherwise.

Video encoding generally refers to the processing of the sequence of pictures that form a video or video sequence. Instead of the term "picture", the terms "frame" or "image" can be used interchangeably in the field of video encoding. Video encoding (or encoding in general) contains two parts: video encoding and video decoding. Video encoding is performed at the source side, typically involving processing (eg, through compression) of the source video images to reduce the amount of data required to represent the video images (for more efficient storage and/or transmission). Video decoding is performed on the recipient (destination) side and usually contains reverse processing compared to the encoder for reconstructing video images. Embodiments relating to "encoding" of video pictures (or pictures in general) are to be understood as referring to "encoding" or "decoding" video pictures or corresponding video sequences. The combination of an encoding part and a decoding part is also called CODEC (encoding and decoding).

In the case of lossless video encoding, the original video images can be restored, i.e. The restored video images are of the same quality as the original video images (provided there are no transmission losses or other data loss during storage or transmission). In the case of lossy video encoding, additional compression is performed, for example through quantization, to reduce the amount of data representing video images that cannot be fully reconstructed at the decoder, i.e. the quality of the restored video images is lower or worse compared to the quality of the original video images.

Several video encoding standards belong to the group of "hybrid lossy video codecs" (i.e., they combine spatial and temporal prediction in the sample domain and 2D transform encoding to apply quantization in the transform domain). Each picture of a video sequence is usually divided into a set of non-overlapping blocks, and encoding is usually performed at the block level. In other words, in an encoder, video is typically processed, that is, encoded, at the block (video block) level, for example using spatial (within a picture) prediction and/or temporal (between pictures) prediction to generate a prediction block, subtracting the prediction block from the current block ( block that is currently being processed/to be processed) to obtain a residual block, transform the residual block and quantize the residual block in the transform domain to reduce the amount of data to be transmitted (compressed), whereas in the decoder, reverse processing compared to the encoder is applied to the encoded or a compressed block to reconstruct the current block for presentation. In addition, the encoder duplicates the decoder's processing loop so that they will both generate identical predictions (eg, inter and intra predictions) and/or reconstructions for processing, i.e., encoding subsequent blocks.

In the following embodiments, the video encoding system 10, the video encoder 20 and the video decoder 30 are described based on FIGS. from 1 to 3.

Fig. 1A is a block diagram showing an exemplary encoding system 10, such as video encoding system 10 (or encoding system 10 for short), which may use the techniques of the present application. Video encoder 20 (or encoder 20 for short) and video decoder 30 (or decoder 30 for short) of video encoding system 10 are examples of devices that may be configured to perform methods in accordance with various examples described herein.

As shown in FIG. 1A, encoding system 10 includes a source device 12 configured to provide encoded picture data 21, for example, to a destination device 14 for decoding the encoded picture data 13.

The source device 12 contains an encoder 20 and can additionally, i.e. optionally comprising a picture source 16, a preprocessor (or preprocessing unit) 18, such as a picture preprocessor 18, and a communications interface or communications unit 22.

The picture source 16 may contain or be any type of picture capturing device, such as a camera for capturing a real world picture, and/or any type of picture generating device, such as a computer graphics processor for generating a computer animated picture, or any type of other device for obtaining and /or providing a real-world image, a computer-generated image (e.g., screen content, virtual reality (VR) image), and/or any combination thereof (e.g., augmented reality (AR) image). The source of the picture can be any type of memory or storage that stores any of the above pictures.

In contrast to the preprocessor 18 and the processing performed by the preprocessing unit 18, the picture or picture data 17 may also be referred to as a raw picture or raw picture data 17.

The preprocessor 18 is configured to receive (raw) picture data 17 and perform preprocessing on the picture data 17 to produce a preprocessed image 19 or preprocessed picture data 19. Pre-processing performed by preprocessor 18 may, for example, comprise cropping, color format conversion (eg, RGB to YCbCr), color correction, or noise reduction. It can be understood that the preprocessing unit 18 may be an optional component.

Video encoder 20 is configured to receive pre-processed picture data 19 and provide encoded picture data 21 (further details will be described below, eg based on FIG. 2).

The communication interface 22 of the source device 12 may be configured to receive encoded picture data 21 and transmit this encoded picture data 21 (or any further processed version thereof) over the communication channel 13 to another device, for example, the destination device 14 or any other device. , for saving or immediate restoration.

The recipient device 14 includes a decoder 30 (eg, video decoder 30) and may additionally, i.e. optionally comprising a communication interface or communication unit 28, a post processor 32 (or post processing unit 32) and a display device 34.

The communication interface 28 of the recipient device 14 is configured to receive encoded picture data 21 (or any further processed version thereof), for example, directly from the source device 12 or from any other source, for example, a storage device, for example, an encoded picture data storage device. , and providing the encoded picture data 21 to the decoder 30.

The communication interface 22 and the communication interface 28 may be configured to transmit or receive encoded picture data 21 or encoded data 13 through a direct communication link between the source device 12 and the destination device 14, such as a direct wired or wireless connection, or through a network of any type, such as a wired or wireless network or any combination thereof, or any type of private and public network, or any type of combination thereof.

The communications interface 22 may, for example, be configured to package the encoded picture data 21 into an appropriate format, such as packets, and/or process the encoded picture data using any type of transmission encoding or processing for transmission over a communications link or network.

The communication interface 28, which is analogous to the communication interface 22, may, for example, be configured to receive the transmitted data and process the transmission data using any kind of appropriate decoding or transmission processing and/or decompression to obtain the encoded picture data 21.

Both communication interface 22 and communication interface 28 may be configured as unidirectional communication interfaces, as shown by the arrow for communication channel 13 in FIG. 1A pointing from a source device 12 to a destination device 14, or as bidirectional communication interfaces, and may be configured to, for example, send and receive messages, for example to establish a connection, to confirm and exchange any other information related to the line communication and/or data transmission, for example, transmission of encoded picture data.

The decoder 30 is configured to receive encoded picture data 21 and provide decoded picture data 31 or decoded picture 31 (further details will be described below, for example, based on FIG. 3 or FIG. 5 .

The post-processor 32 of the recipient device 14 is configured to post-process the decoded picture data 31 (also called reconstructed picture data), for example, the decoded picture 31, to obtain post-processed picture data 33, for example, the post-processed picture 33. Post-processing performed by block 32 post-processing may comprise, for example, color format conversion (eg, from YCbCr to RGB), color correction, cropping or resampling, or any other processing, for example, to prepare decoded picture data 31 for display, for example, by display device 34.

The display device 34 of the recipient device 14 is configured to receive post-processed picture data 33 for displaying the picture, for example, to a user or viewer. The display device 34 may be or include any type of display for presenting the reconstructed picture, such as an integrated or external display or monitor. The displays may, for example, comprise liquid crystal displays (LCD), organic light-emitting diode (OLED), plasma displays, projectors, micro-LED displays, liquid crystal on silicon (LCoS), digital light processor (DLP), or any other type of display.

Although Fig. 1A illustrates source device 12 and destination device 14 as separate devices, embodiments of these devices may also include both or both functionality, source device 12 or corresponding functionality and destination device 14 or corresponding functionality. In such embodiments, source device 12 or related functionality and destination device 14 or related functionality may be implemented using the same hardware and/or software, or using separate hardware and/or software, or any combination thereof.

As will be apparent for the specific element based on the description, the presence and (precise) separation of functionality of the various modules or functions in the source device 12 and/or destination device 14, as shown in FIG. 1A, may vary depending on the actual device and applications.

Encoder 20 (eg, video encoder 20) or decoder 30 (eg, video decoder 30) or both encoder 20 and decoder 30 may be implemented through a processing circuit as shown in FIG. 1B, such as one or more microprocessors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), discrete logic, hardware, dedicated video encoding, or any combination thereof. Encoder 20 may be implemented through processing circuit 46 to implement various modules, as discussed with respect to encoder 20 of FIG. 2 and/or any other encoder system or subsystem described in this document. Decoder 30 may be implemented through processing circuit 46 to implement various modules, as discussed with respect to decoder 30 of FIG. 3 and/or any other decoder system or subsystem described herein. The processing circuit may be configured to perform various operations, which will be described below. As shown in FIG. 5, if the techniques are partially implemented in software, the apparatus may store instructions for the software on a suitable non-transitory computer readable storage medium and may execute the instructions in hardware using one or more processors to execute the techniques of this disclosure. Any one of video encoder 20 and video decoder 30 can be integrated as part of a combined encoder/decoder (CODEC) in one device, for example, as shown in Fig. 1B.

Source device 12 and destination device 14 may comprise any of a wide range of devices, including any type of portable or desktop devices, such as laptops or notebook computers, mobile phones, smartphones, tablets or tablet computers, cameras, desktop computers, set-top boxes , televisions, display devices, digital media players, game consoles, video streaming devices (such as content service servers or content delivery servers), broadcast receiving device, broadcast transmitting device, or the like, and may use or dispense with any type of operating system. In some cases, source device 12 and destination device 14 may be equipped for wireless communication. Thus, source device 12 and destination device 14 may be wireless communication devices.

In some cases, video encoding system 10, illustrated in FIG. 1A is an example only, and the methods of the present application can be applied to video encoding settings (such as video encoding or video decoding) that do not necessarily involve transmitting any data between encoding and decoding devices. In other examples, data is retrieved from local memory, streamed over a network, or the like. A video encoding apparatus may encode and store data into memory, and/or a video decoding apparatus may retrieve and decode data from memory. In some examples, encoding and decoding are performed by devices that do not communicate with each other, but simply encode data into memory and/or retrieve and decode data from memory.

For convenience of description, embodiments of the invention are described herein, for example, with reference to High Efficiency Video Coding (HEVC) or Versatile Video Coding (VVC) reference software, a next-generation video encoding standard developed by the Joint Collaboration Group on Video Coding (JCT-VC). ITU-T Video Coding Experts Group (VCEG) and ISO/IEC Moving Picture Experts Group (MPEG). One of ordinary skill in the art will understand that embodiments of the invention are not limited to HEVC or VVC.

Encoder and encoding method

Fig. 2 shows a schematic block diagram of an exemplary video encoder 20 that is configured to implement the methods of the present application. In the example in FIG. 2, video encoder 20 includes an input 201 (or input interface 201), a remainder calculation block 204, a transform processing block 206, a quantization block 208, an inverse quantization block 210 and an inverse transform processing block 212, a restoration block 214, a loop filter block 220, a buffer 230 ( DPB) of decoded pictures, a mode selection unit 260, an entropy encoding unit 270, and an output 272 (or output interface 272). The mode selection block 260 may include an inter prediction block 244, an intra prediction block 254, and a separating block 262. The inter prediction unit 244 may include a motion estimation unit and a motion compensation unit (not shown). The video encoder 20 shown in FIG. 2 may also be called a hybrid video encoder or a video encoder according to a hybrid video codec.

The remainder calculation block 204, transform processing block 206, quantization block 208, mode selection block 260 may relate to the formation of the forward signal path of the encoder 20, while the inverse quantization block 210, inverse transform processing block 212, restoration block 214, buffer 216, loop filter 220, decoded picture buffer (DPB) 230, inter prediction block 244, and intra prediction block 254 may relate to generating a return signal path of the video encoder 20, wherein the return signal path of the video encoder 20 corresponds to the decoder signal path (see video decoder 30 in FIG. 3) . The inverse quantization unit 210, the inverse transform processing unit 212, the reconstruction unit 214, the loop filter 220, the decoded picture buffer (DPB) 230, the inter prediction unit 244, and the intra prediction unit 254 are also relevant to forming the “on-chip decoder” of the video encoder 20.

Pictures and picture division (pictures and blocks)

Encoder 20 may be configured to receive, for example, through input 201, a picture 17 (or picture data 17), for example a picture from a sequence of pictures forming a video or video sequence. The received picture or picture data may also be a preprocessed picture 19 (or preprocessed picture data 19). For simplicity, the following description refers to picture 17. Picture 17 may also be referred to as the current picture or the picture that is being encoded (particularly in video encoding, to distinguish the current picture from other pictures, for example, previously encoded and/or decoded pictures of the same video sequence, i.e. a video sequence that also contains the current picture).

A (digital) picture is, or can be viewed as, a two-dimensional array or matrix of samples with intensity values. A selection in an array may also be referred to as a pixel (short form of picture element) or pel. The number of samples in the horizontal and vertical direction (or axis) of the array or picture determines the size and/or resolution of the picture. To represent color, three color components are usually used, i.e. the picture can be presented or include three arrays of samples. In RBG format, or color space, a picture contains a corresponding array of red, green, or blue samples. However, in video encoding, each pixel is typically represented in a luminance-chrominance format or color space, such as YCbCr, which contains a luminance component denoted Y (sometimes L is also used instead) and two chrominance components denoted Cb and Cr. The luminance component Y (or luma for short) represents the brightness or intensity of the gray level (eg, as in a grayscale image), while the two chrominance components Cb and Cr represent chrominance or color information components. Accordingly, an image in YCbCr format contains an array of luminance samples with luminance sample values (Y) and two chrominance sample arrays with chrominance values (Cb and Cr). Pictures in RGB format can be converted or converted to YCbCr format and vice versa, the process is also known as color conversion or conversion. If the picture is monochrome, it can only contain an array of brightness samples. Accordingly, a picture can be, for example, an array of luminance samples in a monochrome format or an array of luminance samples and two corresponding arrays of chrominance samples in 4:2:0, 4:2:2 and 4:4:4 color formats.

Embodiments of video encoder 20 may include a picture splitter (not shown in FIG. 2) configured to divide picture 17 into multiple (usually non-overlapping) picture blocks 203. These blocks may also be called root blocks, macroblocks (H.264/AVC) or coding tree blocks (CTB) or coding tree units (CTU) (H.265/HEVC and VVC). The picture division block may be configured to use the same block size for all pictures of a video sequence and a corresponding grid defining the block size, or to vary the block size between pictures or subsets or groups of pictures and divide each picture into appropriate blocks.

In additional embodiments, the video encoder may be configured to receive directly a block 203 of a picture 17, such as one, more, or all of the blocks forming a picture 17. The picture block 203 may also be referred to as a current picture block or a picture block to be encoded.

Like picture 17, picture block 203 is again, or can be thought of as, a two-dimensional array or matrix of samples with intensity values (sample values), albeit smaller in size than picture 17. In other words, block 203 may contain, for example, a single array of samples (e.g. , a brightness array in the case of a monochrome picture 17 or a brightness or chrominance array in the case of a color picture) or three arrays of samples (for example, brightness and two chrominance arrays in the case of color picture 17) or any other number and/or type of arrays depending on the color used format. The number of samples in the horizontal and vertical direction (or axis) of block 203 determines the size of block 203. Accordingly, the block may be, for example, an MxN array of discrete samples (M-columns by N-rows) or an MxN array of transform coefficients.

Embodiments of video encoder 20 shown in FIG. 2 may be configured to encode picture 17 block by block, for example, encoding and prediction is performed for each block 203.

Embodiments of video encoder 20, as shown in FIG. 2 may be further configured to divide and/or encode a picture using slices (also referred to as video slices), wherein the picture may be divided into or encoded using one or more slices (usually non-overlapping), and each slice may contain one or more units (for example, CTU).

Embodiments of video encoder 20 shown in FIG. 2 may be further configured to divide and/or encode a picture using tile groups (also called video tile groups) and/or tiles (also called video tiles), wherein the picture can be divided into or encoded using one or more groups of tiles (usually non-overlapping), and each group of tiles may contain, for example, one or more blocks (for example, CTU) or one or more tiles, each tile, for example, may be rectangular in shape and may contain one or more blocks (eg CTUs), such as full or fractional blocks.

Calculating the remainder

Residual calculation block 204 may be configured to calculate residual block 205 (also referred to as residual 205) based on picture block 203 and prediction block 265 (additional details about prediction block 265 are provided below), for example, by subtracting the sample values of prediction block 265 from the values samples the picture block 203, sample by sample (pixel by pixel), to obtain a residual block 205 in the sample region.

Conversion

Transform processing unit 206 may be configured to apply a transform, such as a discrete cosine transform (DCT) or a discrete sine transform (DST), to the sample values of the residual block 205 to obtain transform coefficients 207 in the transform domain. The transform coefficients 207 may also be referred to as residual transform coefficients and represent the residual block 205 in the transform domain.

Transform processing unit 206 may be configured to apply DCT/DST integer approximations, such as H.265/HEVC-specific transforms. Compared to the orthogonal DCT transform, such integer approximations are usually scaled by some factor. To preserve the norm of the residual block that is processed by the forward and inverse transformations, additional scaling factors are applied during the transformation process. Scaling factors are typically chosen based on some constraints, e.g., scaling factors are powers of two for shift operations, bit depth of transform factors, tradeoff between accuracy and implementation cost, etc. Specific scaling factors, for example, are specified for inverse transform, e.g. by the inverse transform processing unit 212 (and the corresponding inverse transform, e.g., by the inverse transform processing unit 312 in the video decoder 30), and the corresponding scaling factors for the forward transform, such as by the transform processing unit 206, may be set appropriately in the encoder 20.

Embodiments of video encoder 20 (respectively transform processing unit 206) may be configured to output transform parameters, such as the type of transform or transforms, for example, directly or encoded or compressed via entropy encoding unit 270, so that, for example, video decoder 30 may receive and use conversion parameters for decoding.

Quantization

The quantization unit 208 may be configured to quantize the transform coefficients 207 to obtain quantized coefficients 209, for example, by applying scalar quantization or vector quantization. Quantized coefficients 209 may also be referred to as quantized transform coefficients 209 or quantized residual coefficients 209.

The quantization process may reduce the bit depth associated with some or all of the transform coefficients 207. For example, an n-bit transform factor may be rounded to an m-bit transform factor during quantization, where n is greater than m. The degree of quantization can be modified by adjusting the quantization parameter (QP). For example, for scalar quantization, different scaling may be applied to achieve finer or coarser quantization. Smaller quantization step sizes correspond to finer quantization, while larger quantization step sizes correspond to coarser quantization. The applicable quantization step size may be specified by the quantization parameter (QP). The quantization parameter may, for example, be an index to a predefined set of applicable quantization step sizes. For example, small quantization parameters may correspond to fine quantization (small quantization step sizes), and large quantization parameters may correspond to coarse quantization (large quantization step sizes), or vice versa. Quantization may include dividing by a quantization step size, and corresponding and/or inverse dequantization, such as by inverse quantization unit 210, may include multiplying by a quantization step size. Embodiments in accordance with some standards, such as HEVC, may be configured to use a quantization parameter to determine the quantization step size. Typically, the quantization step size can be calculated based on the quantization parameter using a fixed-point approximation of the equation including division. Additional scaling factors can be introduced for quantization and dequantization to recover the residual block norm that may have been changed due to the scaling used in the fixed point approximation of the above equation for the quantization step size and the quantization parameter. In one exemplary implementation, inverse transform scaling and dequantization may be combined. Alternatively, customized quantization tables can be used and signaled from encoder to decoder, for example in a bit stream. Quantization is a lossy operation in which losses increase as the size of the quantization steps increases.

Embodiments of video encoder 20 (respectively quantization block 208) may be configured to output quantization parameters (QPs), for example, directly or encoded through entropy encoding block 270, so that, for example, video decoder 30 can receive and apply the quantization parameters for decoding.

Inverse quantization

The inverse quantization block 210 is configured to apply the inverse quantization of the quantization block 208 to the quantized coefficients to obtain the dequantized coefficients 211, for example, by applying an inverse quantization scheme to that applied by the quantization block 208, based on or using the same quantization step size as quantization unit 208. The dequantized coefficients 211 may also be referred to as the dequantized residual coefficients 211 and correspond - although they are typically not identical to the transform coefficients due to quantization loss - to the transform coefficients 207.

Reverse conversion

The inverse transform processing unit 212 is configured to apply an inverse transform relative to the transform applied by the transform processing unit 206, such as an inverse discrete cosine transform (DCT) or an inverse discrete sine transform (DST) or other inverse transforms to obtain the reconstructed residual block 213 (or corresponding dequantized coefficients 213) in the sample area. The reconstructed residual block 213 may also be referred to as a transform block 213.

Recovery

The reconstruction block 214 (e.g., add block or adder 214) is configured to add the transform block 213 (i.e., the reconstructed residual block 213) with the prediction block 265 to obtain the reconstructed block 215 in the sample domain, for example, through sample-by-sample addition the sample values of the reconstructed residual block 213 and the sample values of the prediction block 265.

Filtration

Loop filter block 220 (or “loop filter” 220 for short) is configured to filter the reconstructed block 215 to obtain a filtered block 221, or more generally to filter the reconstructed samples to obtain filtered samples. The edge filter block, for example, is configured to smooth pixel transitions or otherwise improve video quality . Loop filter block 220 may include one or more loop filters, such as a deblocking filter, a sample adaptive offset (SAO) filter, or one or more other filters, such as a two-way filter, an adaptive loop filter (ALF), anti-aliasing filters, boost filters, etc. sharpening or joint filters, or any combination of them. Although the loop filter block 220 is shown in FIG. 2 as a loop filter, in other configurations the loop filter block 220 may be implemented as a post-loop filter. Filtered block 221 may also be referred to as filtered reconstructed block 221.

Embodiments of video encoder 20 (respectively, loop filter block 220) may be configured to output loop filter parameters (such as sample-adaptive offset information), for example, directly or encoded through entropy encoding block 270, so that, for example, decoder 30 can receive and apply the same loop filter parameters or corresponding loop filters for decoding.

Buffer of decoded pictures

Decoded picture buffer (DPB) 230 may be a memory that stores reference pictures, or generally reference picture data, for encoding video data by video encoder 20. DPB 230 may be formed by any of a variety of storage devices, such as dynamic random access memory (DRAM). DRAM), including synchronous DRAM (SDRAM), magnetoresistive RAM (MRAM), resistive RAM (RRAM), or other types of memory devices. The decoded picture buffer (DPB) 230 may be configured to store one or more filtered blocks 221. The decoded picture buffer 230 may be further configured to store other previously filtered blocks, such as previously reconstructed and filtered blocks 221, the same current picture, or different pictures, for example, previously restored pictures, and can provide complete previously restored ones, i.e. decoded, pictures (and corresponding reference blocks and samples) and/or partially reconstructed current picture (and corresponding reference blocks and samples), for example, for inter-prediction. The decoded picture buffer (DPB) 230 may also be configured to store one or more unfiltered reconstructed blocks 215, or generally unfiltered reconstructed samples, for example if the reconstructed block 215 is not filtered by the loop filter block 220, or any other further processed version of the reconstructed blocks. or samples.

Mode selection (separation and prediction)

The mode selection unit 260 includes a division unit 262, an inter prediction unit 244, and an intra prediction unit 254, and is configured to receive or obtain original picture data, such as the original block 203 (current block 203 of the current picture 17), and reconstructed picture data, such as filtered and/or unfiltered reconstructed samples or blocks of the same (current) picture and/or from one or a plurality of previously decoded pictures, such as from decoded picture buffer 230 or other buffers (eg, a line buffer, not shown). The reconstructed picture data is used as reference picture data for prediction, for example, inter-prediction or intra-prediction to obtain a prediction block 265 or a predictor 265.

The mode selector 260 may be configured to determine or select a split for the current block prediction mode (including no split) and a prediction mode (eg, intra or inter prediction mode) and generate a corresponding prediction block 265 that is used to calculate the residual block 205 and to restore the recovered block 215.

Embodiments of the mode selector 260 may be configured to select a split and prediction mode (e.g., from those supported by or available to the mode selector 260) that provides the best match, or in other words, the minimum remainder (the minimum remainder means the best compression for transmission or storage), or minimal signaling overhead (minimum signaling overhead means better compression for transmission or storage), or which takes into account or balances both factors. Mode selector 260 may be configured to determine a division and prediction mode based on rate distortion optimization (RDO), that is, selecting a prediction mode that provides minimal rate distortion. Terms such as "best", "minimal", "optimal", etc. in this context do not necessarily refer to the all-encompassing "best", "minimum", "optimal", etc., but may also refer to meeting a selection or termination criterion, such as when a value exceeds or falls below a threshold, or other constraints, potentially leading to "suboptimal selection" but reducing complexity and processing time.

In other words, partitioning unit 262 may be configured to partition block 203 into smaller block sections or sub-blocks (which form blocks again), for example, iteratively using quad-tree (QT) partitioning, binary partitioning (BT), or ternary tree partitioning ( TT) or any combination thereof, and performing, for example, prediction for each of the block sections or subblocks, wherein the mode selection comprises selecting a tree structure of the shared block 203, and the prediction modes are applied to each of the block sections or subblocks.

Next, the division (eg, through the division module 260) and prediction processing (via the inter prediction module 244 and the intra prediction module 254) performed by the exemplary video encoder 20 are explained in more detail.

Separation

Divider 262 may divide (or split) the current block 203 into smaller sections, such as smaller blocks of square or rectangular size. These smaller blocks (which may also be called subblocks) can be further divided into even smaller sections. This is also called tree partitioning or hierarchical tree partitioning, in which the root block, for example at root level 0 of the tree (hierarchy level 0, depth 0), can be recursively partitioned, for example divided into two or more blocks of the next lower tree level, such as nodes at tree level 1 (hierarchy level 1, depth 1), whereby these blocks can be again divided into two or more blocks of the next lower level, such as tree level 2 (hierarchy level 2, depth 2), etc. until the split will not complete, for example because a termination criterion has been met, such as reaching the maximum tree depth or minimum block size. Blocks that are not further divided are also called leaf blocks or leaf nodes of a tree. A tree using two partitions is called a binary tree (BT), a tree using three partitions is called a ternary tree (TT), and a tree using four partitions is called a quadtree (QT).

As mentioned earlier, the term "block" as used herein can be a part, particularly a square or rectangular part of a picture. With respect to, for example, HEVC and VVC, a block may be or correspond to a coding tree unit (CTU), a coding unit (CU), a prediction unit (PU) and a transform unit (TU) and/or corresponding blocks, for example a coding tree block ( CTB), coding block (CB), transform block (TB) or prediction block (PB).

For example, a coding tree unit (CTU) may be or contain a CTB of luma samples, two corresponding CTBs of chroma samples of an image that has three arrays of samples, or a CTB of samples of a monochrome image or an image that is encoded using three separate color planes and the syntax structures used for encoding samples. Accordingly, a coding tree unit (CTB) may be a block of NxN samples for some value of N, so that dividing a component by the CTB is a division. A coding unit (CU) may be or comprise a luma sample encoding unit, two corresponding chroma sample encoding units of a picture that has three sample arrays, or a monochrome sample encoding unit or a picture that is encoded using three separate color planes and syntactic structures used for encoding samples. Accordingly, a coding block (CB) may be a block of MxN samples for some values of M and N, so that dividing the CTB into coding blocks is a division.

In embodiments, for example, according to HEVC, a coding tree unit (CTU) can be partitioned into CUs using a quadtree structure designated as a coding tree. The decision of whether to encode a picture region using inter-picture (temporal) or intra-picture (spatial) prediction is made at the CU level. Each CU can be further partitioned into one, two or four PUs according to the PU partition type. Within a single PU, the same prediction process is applied, and the relevant information is passed to the PU-based decoder. After obtaining the residual block by applying a prediction process based on the partition type on the PU, the CU can be divided into transformation units (TUs) according to another quadtree structure similar to the encoding tree for the CU.

In embodiments, for example, in accordance with the newest video encoding standard currently under development, called Versatile Video Coding (VVC), for example, a combined quadtree and binary tree (QTBT) partitioning is used to divide the encoding block. In a block structure, a QTBT CU can be either square or rectangular in shape. For example, a coding tree unit (CTU) is first divided into a quadtree structure. The leaf nodes of a quadtree are further separated by a binary tree or ternary (or ternary) tree structure. The leaf nodes of a partitioning tree are called coding units (CUs), and this partitioning is used to process prediction and transformation without any further partitioning. This means that CU, PU and TU have the same block size in the QTBT encoding block structure. In parallel, multiple partitioning, such as ternary tree partitioning, can be used in conjunction with the QTBT block structure.

In one example, the mode selector 260 of the video encoder 20 may be configured to perform any combination of the division methods described herein.

As described above, video encoder 20 is configured to determine or select the best or optimal prediction mode from a set of (eg, predefined) prediction modes. The set of prediction modes may comprise, for example, intra prediction modes and/or inter prediction modes.

Internal prediction

The set of intra prediction modes may contain 35 different intra prediction modes, e.g., non-directional modes such as DC (or average) mode and planar mode, or directional modes, such as those defined in HEVC, or may contain 67 different intra prediction modes, e.g. non-directional modes such as DC (or average) mode and planar mode, or directional modes such as those defined in VVC.

The intra prediction block 254 is configured to use reconstructed discrete samples of neighboring blocks of the same current picture to generate an intra prediction block 265 according to the intra prediction mode of the set of intra prediction modes.

The intra prediction block 254 (or more generally the mode selector 260) is further configured to output intra prediction parameters (or more generally information indicating the selected intra prediction mode for the block) to the entropy encoding block 270 in the form of syntax elements 266 for including pictures in the data 21 so that, for example, the video decoder 30 can receive and use the prediction parameters for decoding.

External prediction

The set of (candidate) inter prediction modes depends on the available reference pictures (i.e., previous at least partially decoded pictures, such as those stored in DBP 230) and other inter prediction parameters, such as whether the entire reference picture is used or only a portion, e.g. search window area around the area of the current block, a reference picture to find the most suitable reference block and/or, for example, whether pixel interpolation, such as half/semi-pel and/or quarter-pel interpolation, is applied or not .

In addition to the above prediction modes, a skip mode and/or a direct mode may be used.

The inter-prediction block 244 may include a motion estimation (ME) block and a motion compensation (MC) block (both not shown in FIG. 2). The motion estimator may be configured to receive or obtain a picture block 203 (current picture block 203 for the current picture 17) and a decoded picture 231, or at least one or more previously reconstructed blocks, such as reconstructed blocks of one or many different/different previously 231 decoded pictures for motion estimation. For example, the video sequence may contain the current picture and previously decoded pictures 231 or, in other words, the current picture and previously decoded pictures 231 may be part of or form a sequence of pictures making up the video sequence.

The encoder 20 may, for example, be configured to select a reference block from a plurality of reference blocks of the same or different pictures from a plurality of other pictures and provide a reference picture (or reference picture index) and/or an offset (spatial offset) between the position (x, y coordinates ) of the reference block and the position of the current block as inter-prediction parameters to the motion estimation block. This displacement is also called motion vector (MV).

The motion compensation unit is configured to obtain, for example, receiving an inter prediction parameter and performing inter prediction based on or using the inter prediction parameter to obtain an inter prediction unit 265. The motion compensation performed by the motion compensation block may include obtaining or generating a prediction block based on a motion vector/block determined by motion estimation, possibly performing interpolations with sub-pixel precision. Interpolation filtering can generate additional pixel samples from known pixel samples, thereby potentially increasing the number of candidate prediction blocks that can be used to encode a picture block. After receiving the motion vector for the PU block of the current picture, the motion compensation unit can determine the location of the prediction block pointed to by the motion vector in one of the reference picture lists.

The motion compensation block may also generate syntax elements associated with blocks and video slices for use by video decoder 30 when decoding the picture blocks of that video slice. In addition to or as an alternative to slices and corresponding syntax elements, groups of tiles and/or tiles and corresponding syntax elements may be generated or used.

Entropy encoding

The entropy encoding block 270 is configured to employ, for example, an entropy encoding algorithm or scheme (e.g., a variable length coding (VLC) scheme, a context adaptive VLC (CAVLC) scheme, an arithmetic encoding scheme, binarization, a context adaptive binary arithmetic encoding ( CABAC) based on the syntax of context-adaptive binary arithmetic coding (SBAC), probability interval entropy encoding (PIPE), or other entropy encoding method or methodology) or bypass (uncompressed) with respect to quantized coefficients 209, inter-prediction parameters, parameters intra prediction, loop filter parameters, and/or other syntax elements to obtain encoded picture data 21, which may be output through output 272, for example in the form of an encoded bit stream 21, so that, for example, video decoder 30 can receive and use these parameters for decoding. The encoded bit stream 21 may be transmitted to video decoder 30 or stored in memory for later transmission or retrieval by video decoder 30.

Other changes to the structure of video encoder 20 may be used to encode the video stream. For example, non-transform encoder 20 may quantize the residual signal directly without transform processing unit 206. In another implementation, encoder 20 may have a quantization unit 208 and an inverse quantization unit 210 combined into a single unit.

Decoder and decoding method

Fig. 3 shows an example of a video decoder 30 that is configured to implement the methods of the present application. The video decoder 30 is configured to receive encoded picture data 21 (e.g., a coded bit stream 21), such as encoded by an encoder 20, to obtain a decoded picture 331. The encoded picture data or bit stream contains information for decoding the encoded picture data, e.g., data that represent picture blocks of an encoded video slice (and/or tiles or groups of tiles) and associated syntax elements.

In the example in FIG. 3, decoder 30 includes an entropy decoding block 304, an inverse quantization block 310, an inverse transform processing block 312, a recovery block 314 (e.g., adder 314), a loop filter 320, a decoded picture buffer (DPB) 330, a mode applying block 360, a block 344 inter prediction and intra prediction block 354. The inter prediction unit 344 may be or include a motion compensation unit. Video decoder 30 may, in some examples, perform a decoding step that is generally the reverse of the encoding step described with respect to video encoder 100 in FIG. 2.

As described with respect to the encoder 20, the inverse quantization unit 210, the inverse transform processing unit 212, the reconstruction unit 214, the loop filter 220, the decoded picture buffer (DPB) 230, the inter-prediction unit 344, and the intra-prediction unit 354 also relate to forming the “on-chip decoder”. » video encoder 20. Accordingly, the inverse quantization block 310 may be identical in function to the inverse quantization block 110, the inverse transform processing block 312 may be identical in function to the inverse transform processing block 212, the restoration block 314 may be identical in function to the reconstruction block 214, loop filter 320 may be identical in function to loop filter 220, and decoded picture buffer 330 may be identical in function to decoded picture buffer 230. Therefore, the explanations provided for the corresponding blocks and functions of the video encoder 20 apply correspondingly to the corresponding blocks and functions of the video decoder 30.

Entropy Decoding

Entropy decoding block 304 is configured to parse the bit stream 21 (or generally picture data 21) and perform, for example, entropy decoding on picture data 21 to obtain, for example, quantized coefficients 309 and/or decoded encoding parameters (not shown in FIG. 3), for example, any or all of the inter prediction parameters (eg, reference picture index and motion vector), intra prediction parameter (eg, index or intra prediction mode), transform parameters, quantization parameters, loop filter parameters, and /or other syntax elements. Entropy decoding unit 304 may be configured to apply decoding algorithms or schemes corresponding to encoding schemes as described with respect to entropy encoding unit 270 of encoder 20. Entropy decoding unit 304 may be further configured to provide inter prediction parameters, an intra prediction parameter, and/or or other syntax elements to block 360 for applying the mode and other parameters to other blocks of decoder 30. Video decoder 30 may receive syntax elements at the video slice level and/or video block level. In addition to or as an alternative to slices and corresponding syntax elements, groups of tiles and/or tiles and corresponding syntax elements may be received and/or used.

Inverse quantization

The inverse quantization unit 310 may be configured to receive quantization parameters (QPs) (or, more generally, information related to inverse quantization) and quantized coefficients from the encoded picture data 21 (e.g., by parsing and/or decoding, e.g., by entropy decoding unit 304) and applying, based on the quantization parameters, inverse quantization on the decoded quantized coefficients 309 to obtain dequantized coefficients 311, which may also be referred to as transform coefficients 311. The inverse quantization process may include using a quantization parameter determined by video encoder 20 for each video block in a video slice (or tile or group of tiles) to determine the degree of quantization and, likewise, the degree of inverse quantization to be applied.

Reverse conversion

The inverse transform processing unit 312 may be configured to receive the dequantized coefficients 311, also referred to as transform coefficients 311, and apply a transform to the dequantized coefficients 311 to obtain reconstructed residual blocks 213 in the sample region. The reconstructed residual blocks 213 may also be referred to as transform blocks 313. The conversion may be an inverse transform, such as an inverse DCT, an inverse DST, an integer integer transform, or a conceptually similar inverse transform process. The inverse transform processing unit 312 may be further configured to receive transform parameters or corresponding information from the encoded picture data 21 (e.g., by parsing and/or decoding, e.g., by entropy decoding unit 304) to determine a transform to be applied to dequantized coefficients 311.

Recovery

The reconstruction block 314 (e.g., add block or adder 314) may be configured to add the reconstructed residual block 313 with the prediction block 365 to obtain a reconstructed block 315 in the sample domain, for example by adding the sample values of the reconstructed residual block 313 and the sample values of the block 365 predictions.

Filtration

A loop filter block 320 (either in the encoding loop or after the encoding loop) is configured to filter the reconstructed block 315 to produce a filtered block 321, for example, to smooth pixel transitions or otherwise improve video quality. Loop filter block 320 may include one or more loop filters, such as a deblocking filter, a sample-adaptive offset (SAO) filter, or one or more other filters, such as a two-way filter, an adaptive loop filter (ALF), anti-aliasing filters, enhancement filters, etc. sharpening or joint filters, or any combination of them. Although the loop filter block 320 is shown in FIG. 3 as a loop filter, in other configurations the loop filter block 320 may be implemented as a post-loop filter.

Buffer of decoded pictures

The decoded video picture blocks 321 are then stored in a decoded picture buffer 330, which stores the decoded pictures 331 as reference pictures for subsequent motion compensation for other pictures and/or for output or display.

The decoder 30 is configured to output the decoded picture 311, for example, through an output 312 for presentation or viewing to a user.

Prediction

The inter prediction block 344 may be identical to the inter prediction block 244 (in particular, the motion compensation block), and the intra prediction block 354 may be identical to the inter prediction block 254 in function, and makes splitting or splitting decisions and performs prediction based on the splitting parameters and /or prediction or corresponding information received from the encoded picture data 21 (eg, by parsing and/or decoding, eg, by entropy decoding unit 304). The mode application block 360 may be configured to perform prediction (intra- or inter-prediction) for each block based on the reconstructed images, blocks, or corresponding samples (filtered or unfiltered) to obtain a prediction block 365.

When a video slice is encoded as an intra-encoded (I) slice, an intra prediction block 354 of a mode applying block 360 is configured to generate a prediction block 365 for a picture block of the current video slice based on the signaled intra prediction mode and data from previously decoded current picture blocks. When a video picture is encoded as an externally encoded (i.e., B or P) slice, an inter-prediction block 344 (e.g., a motion compensation block) of the mode applying block 360 is configured to create prediction blocks 365 for the video block of the current video slice based on motion vectors and other elements syntax received from entropy decoding block 304. For inter-prediction, prediction blocks may be created from one of the reference pictures within one of the reference picture lists. Video decoder 30 may construct reference picture lists, List 0 and List 1, using default construction methods based on reference pictures stored in DPB 330. The same or similar may be used for or through embodiments using tile groups (e.g., video tile groups ) and/or tiles (eg, video tiles) in addition to or alternatively to slices (eg, video slices), for example, video may be encoded using groups of I, P or B tiles and/or tiles.

The mode applying unit 360 is configured to determine prediction information for a video block of the current video slice by parsing motion vectors or related information and other syntax elements, and uses the prediction information to create prediction blocks for the current video block to be decoded. For example, mode apply block 360 uses some of the received syntax elements to determine the prediction mode (e.g., intra or inter prediction) used to encode video blocks of a video slice, type of inter prediction slice (e.g., B-slice, P-slice, or GPB slice) , construction information for one or more reference picture lists for a slice, motion vectors for each externally encoded video block of the slice, inter-prediction status for each externally encoded video block of the slice, and other information for decoding video blocks in the current video slice. The same or similar may apply to or through embodiments using tile groups (eg, video tile groups) and/or tiles (eg, video tiles) in addition to or alternatively to slices (eg, video slices), for example, video may be encoded with using groups of I, P or B tiles and/or tiles.

Embodiments of video decoder 30, as shown in FIG. 3 may be configured to split and/or decode a picture using slices (also referred to as video slices), wherein the picture may be divided into or encoded using one or more slices (usually non-overlapping), and each slice may contain one or more blocks (for example, CTU).

Embodiments of video decoder 30 shown in FIG. 3 may be configured to divide and/or decode a picture using groups of tiles (also called video tile groups) and/or tiles (also called video tiles), wherein the picture can be divided into or decoded using one or more tile groups ( typically non-overlapping), and each group of tiles may contain, for example, one or more blocks (eg, CTU) or one or more tiles, where each tile, for example, may be rectangular in shape and may contain one or more blocks (eg, CTU ), such as full or fractional blocks.

Other embodiments of the video decoder 30 may be used to decode the encoded picture data 21. For example, decoder 30 may produce an output video stream without loop filtering block 320. For example, the non-transform decoder 30 may perform inverse quantization of the residual signal directly without detransforming processing unit 312 for some blocks or frames. In another implementation, video decoder 30 may have an inverse quantization block 310 and an inverse transform processing block 312 combined into one block.

It should be understood that in the encoder 20 and decoder 30, the result of processing of some current stage may be further processed and then output to the next stage. For example, after interpolation filtering, motion vector acquisition, or contour filtering, an additional operation such as Clip or offset may be performed on the result of the interpolation filtering, motion vector acquisition, or contour filtering processing.

It should be noted that additional operations may be applied to the resulting current block motion vectors (including, but not limited to, affine mode checkpoint motion vectors, affine mode, planar, ATMVP mode subblock motion vectors, temporary motion vectors, and the like). For example, the value of a motion vector is limited to a predefined range according to its representing bit. If the representing bit of the motion vector is bitDepth, then the range is -2^(bitDepth-1) ~ 2^(bitDepth-1)-1, where "^" means exponentiation. For example, if bitDepth is 16, the range is -32768~32767; if bitDepth is 18, the range is -131072~131071. For example, the value of the resulting motion vector (for example, the MV of four 4x4 subblocks in one 8x8 block) is limited so that the maximum difference between the entire parts of the MV of four 4x4 subblocks does not exceed N pixels, for example, is no more than 1 pixel . Here are two ways to constrain the motion vector according to bitDepth.

Method 1: Removing MSB (Most Significant Bit) Overflow via Streaming Operations

ux=(mvx+2 ^bitDepth ) % 2 ^bitDepth (1)

mvx=( ux >= 2 ^bitDepth-1 ) ? (ux - 2 ^bitDepth ): ux (2)

uy=(mvy+2 ^bitDepth ) % 2 ^bitDepth (3)

mvy=( uy >= 2 ^bitDepth-1 ) ? (uy - 2 ^bitDepth ): uy (4)

where mvx represents the horizontal component of the motion vector of the image block or sub-block, mvy represents the vertical component of the motion vector of the image block or sub-block, and ux and uy indicate an intermediate value;

For example, if the value of mvx is -32769, after applying formulas (1) and (2), the resulting value is 32767. In a computer system, decimal numbers are stored as two's complement. The two's complement for -32769 is 1.0111,1111,1111, 1111 (17 bits), then the MSB is discarded, so the resulting two's complement is 0111,1111,1111,1111 (decimal is 32767), which is the same as the output from the application of formulas (1) and (2).

ux=(mvpx + mvdx +2 ^bitDepth ) % 2 ^bitDepth (5)

mvx=( ux >= 2 ^bitDepth-1 ) ? (ux - 2 ^bitDepth ): ux (6)

uy=(mvpy + mvdy +2 ^bitDepth ) % 2 ^bitDepth (7)

mvy=( uy >= 2 ^bitDepth-1 ) ? (uy - 2 ^bitDepth ): uy (8)

Operations can be applied during the summation of mvp and mvd, as shown in formulas (5) to (8).

Method 2: Remove MSB overflow by truncating the value

vx=Clip3(-2 ^bitDepth-1 , 2 ^bitDepth-1 -1, vx)

vy=Clip3(-2 ^bitDepth-1 , 2 ^bitDepth-1 -1, vy)

where vx represents the horizontal component of the motion vector of the image block or sub-block, vy represents the vertical component of the motion vector of the image block or sub-block; x, y and z respectively correspond to the three input values of the MV clipping process, and the definition of the Clip3 function is as follows:

Clip3( x, y, z ) =

Fig. 4 is a schematic diagram of a video encoding apparatus 400 according to an embodiment of the present disclosure. The video encoding apparatus 400 is suitable for implementing the disclosed embodiments that are described herein. In an embodiment, video encoding device 400 may be a decoder, such as video decoder 30 of FIG. 1A, or an encoder such as video encoder 20 of FIG. 1A.

The video encoding device 400 includes input ports 410 (or input ports 410) and receiver (Rx) blocks 420 for receiving data; a processor, logical unit, or central processing unit (CPU) 430 for processing data; transmitter (Tx) blocks 440 and output ports 450 (or output ports 450) for data transmission; and a memory 460 for storing data. Video encoding device 400 may also include optical-to-electrical (OE) components and electrical-to-optical (EO) components connected to input ports 410, receiver blocks 420, transmitter blocks 440, and output ports 450 to provide optical input or output. or electrical signals.

The processor 430 is implemented in hardware and software. Processor 430 may be implemented as one or more CPU chips, cores (eg, a multi-core processor), FPGA, ASIC, and DSP. Processor 430 communicates with input ports 410, receiver blocks 420, transmitter blocks 440, output ports 450, and memory 460. Processor 430 includes an encoding block 470. Encoding block 470 implements the disclosed embodiments described above. For example, encoding block 470 implements, processes, prepares, or provides various encoding operations. Therefore, the inclusion of the encoding module 470 provides a significant improvement in the functionality of the video encoding device 400 and allows the video encoding device 400 to be transformed into a different state. Alternatively, encoding block 470 is implemented as instructions stored in memory 460 and executed by processor 430.

Memory 460 may include one or more disks, tape drives, and solid-state drives and may be used as an overflow storage device for storing programs when such programs are selected for execution and for storing instructions and data that are read during execution of the programs. Memory 460 may be, for example, volatile and/or non-volatile and may be read only memory (ROM), random access memory (RAM), ternary memory associative memory (TCAM), and/or static random access memory (SRAM).

Fig. 5 is a simplified block diagram of a device 500 that can be used as one or both of the source device 12 and the destination device 14 of FIG. 1 according to an exemplary embodiment .

The processor 502 in the device 500 may be a central processing unit. Alternatively, processor 502 may be any other type of device or plurality of devices capable of manipulating or processing information currently existing or developed in the future. Although the disclosed implementations may be practiced with a single processor, as shown, for example, with processor 502, speed and efficiency benefits may be achieved by using more than one processor.

Memory 504 in device 500 may be read-only memory (ROM) or random access memory (RAM) in an implementation. Memory 504 may be any other suitable type of storage device. Memory 504 may include code and data 506 accessed by processor 502 using bus 512. Memory 504 may further include an operating system 508 and application programs 510, wherein application programs 510 include at least one program which allows the processor 502 to perform the methods described herein. For example, application programs 510 may include applications 1 through N, which further include a video encoding application that performs the methods described herein.

Device 500 may also include one or more output devices, such as a display 518. Display 518 may be, in one example, a touch display that combines a display with a touch element capable of sensing touch inputs. Display 518 may be coupled to processor 502 via bus 512.

Although depicted here as a single bus, bus 512 of device 500 may be composed of multiple buses. In addition, secondary storage 514 may be directly coupled to other components of device 500 or may be accessible over a network and may comprise a single plug-in unit, such as a memory card, or multiple units, such as multiple memory cards. Thus, device 500 can be implemented in a wide variety of configurations.

Motion Vector Refinement (MVR)

Motion vectors are typically at least partially determined on the encoder side and transmitted to the decoder in the encoded bitstream. However, motion vectors can also be refined at the decoder (as well as at the encoder), starting from the original motion vectors specified in the bitstream. In such a case, for example, the similarity between fragments of already decoded pixels indicated by the original motion vectors can be used to improve the accuracy of the original motion vectors. This motion refinement provides the benefit of reducing signaling overhead: the accuracy of the original motion is improved equally in both the encoder and decoder, and thus no additional signaling is required for refinement.

It should be noted that the original motion vectors before refinement may not be the best motion vectors that lead to the best prediction. Since the original motion vectors are signaled in a bitstream, it may not be possible to represent the original motion vector with very high precision (which would increase the bit rate), so the motion vector refinement process is used to improve the original motion vector. The source motion vectors may, for example, be motion vectors that are used in predicting a neighbor block of the current block. In this case, it is sufficient to signal an indication in the bitstream indicating which neighboring block's motion vectors are being used by the current block. This prediction mechanism is very effective in reducing the number of bits to represent the original motion vectors. However, the accuracy of the original motion vectors may be low because, in general, the motion vectors of two adjacent blocks are not expected to be identical.

To further improve the accuracy of the motion vectors without further increasing signaling overhead, it may be useful to further refine the motion vectors derived at the encoder side and provided (signaled) in the bitstream. Motion vector refinement can be performed in the decoder without the help of an encoder. The encoder in its decoder loop can use the same refinement to obtain corresponding refined motion vectors that would be available at the decoder. Refinement for the current block that is reconstructed in the current picture is performed by determining a pattern of reconstructed samples, defining a search space around the original motion information for the current block, and finding in the search space the part of the reference picture that best matches the pattern. The best matching part determines refined motion vectors for the current block, which are then used to obtain inter-predicted samples for the current block, that is, the current block that is being reconstructed.

Motion vector refinement is part of the inter prediction block (244) in FIG. 2 and 344 in Fig. 3.

Refinement of the motion vector can be performed in accordance with the following steps:

Typically, the initial motion vectors can be determined based on the indication in the bit stream. For example, an index may be signaled in the bitstream that indicates a position in a list of candidate motion vectors. In another example, the motion vector predictor index and the motion vector difference value may be signaled in the bitstream. Motion vectors that are determined based on the indication in the bitstream are defined as original motion vectors. In the case of bidirectional prediction, where the outer prediction for the current block is obtained as a weighted combination of the predicted block of samples that are determined according to two motion vectors, let the original motion vector in the first reference picture in the list L0 be designated as MV0; and the original motion vector in the second reference picture in the list L1 is designated as MV1.

Using the original motion vectors, pairs of candidate motion vectors (MVs) for refinement are determined. At a minimum, two pairs of candidates for refinement must be identified. Typically, pairs of candidate motion vectors for refinement are determined based on the original pair of motion vectors (MV0, MV1). In addition, pairs of candidate MVs are determined by adding small motion vector differences to MV0 and MV1. For example, candidate MV pairs could include the following:

(MV0, MV1)

(MV0 + (0,1), MV1 + (0, -1))

(MV0 + (1.0), MV1 + (-1.0))

(MV0 + (0, -1), MV1 + (0,1))

(MV0 + (-1.0), MV1 + (1.0))

...

where (1, -1) denotes a vector that has an offset of 1 in the horizontal (or x) direction and an offset of -1 in the vertical (or y) direction.

It should be noted that the above list of candidate pairs are merely examples for explanation, and the invention is not limited to a specific list of candidates.

Pairs of candidate motion vectors (MVs) for refinement form the search space of the motion vector refinement process.

In bidirectional current block prediction, two prediction blocks obtained using the corresponding first list motion vector L0 and second list motion vector L1 are combined into one prediction signal, which can achieve better adaptation to the original signal than a single prediction, resulting in less residual information and possibly more efficient compression.

In motion vector refinement, two prediction blocks obtained using the corresponding first motion vector and the second motion vector of a pair of candidate MVs are compared based on a similarity metric for each of the pairs of candidate MVs for refinement. The pair of candidate MVs that gives the greatest similarity is usually selected as refined motion vectors. Denoted as MV0' and MV1', the refined motion vector in the first reference picture in the L0 list and the refined motion vector in the second reference picture in the L1 list, respectively. In other words, predictions are obtained according to the motion vector of list L0 and the motion vector of list L1 of pairs of candidate motion vectors, which are then compared based on a similarity metric. The candidate motion vector pair that has the highest associated similarity is selected as the refined MV pair.

Typically, the refinement process results in refined MVs. The refined MVs may be the same as the original MVs, or may be different from the original MVs, depending on which candidate MV pair achieves the highest similarity, the candidate MV pair formed by the original MVs is also among the candidate MV pairs. In other words, if the highest pair of candidate MVs that achieves the highest similarity is formed by the original MVs, the refined MVs and the original MVs are equal to each other.

Instead of choosing a position that maximizes the similarity metric, another method is to choose a position that minimizes the dissimilarity metric. The dissimilarity comparison measure can be SAD (Sum of Absolute Differences), MRSAD (Mean Remote Sum of Absolute Differences, SSE (Sum of Squared Errors), etc. The SAD between two prediction blocks can be obtained using a pair of candidate MVs (CMV0, CMV1), SAD can be calculated as follows:

where nCbH and nCbW are the height and width of the prediction blocks, the function abs(a) specifies the absolute value of the argument a, predSAmplesL0 and predSAmplesL1 are the prediction block samples obtained in accordance with the pair of candidate MVs, which is denoted as (CMV0, CMV1).

Alternatively, dissimilarity comparison measures can be obtained by evaluating only a subset of samples in the prediction block to reduce the amount of computation. Example below where sample rows are alternatively included in the SAD calculation (every other row is evaluated).

One example of motion vector refinement is explained in the document “JVET-M1001-v3, Versatile Video Coding (Draft 4)” of JVET (of ITU-T SG 16 WP 3 and ISO/IEC JTC 1/SC 29/WG 11), which is publicly available at http://phenix.it-sudparis.eu/jvet/.” Section “8.4.3 Decoder side motion vector refinement process” in the document illustrates motion vector refinement.

To reduce internal memory requirements for refinement, in some embodiments, the motion vector refinement process may be performed independently on blocks of luminance samples obtained by dividing an encoded block of samples that exceed some predetermined width or predetermined height in luminance samples into subblocks of samples that are less than or equal to equal to a predefined width and a predefined height in brightness. The refined MV pair for each subblock in a divided coded block may be different. External prediction is then performed for luma and chrominance for each sub-block using the refined MV pair of that sub-block.

Each MV of a pair of source MVs can have fractional pixel precision. In other words, the MV indicates an offset between the current block of samples and the resampled reference region, and this offset may indicate a fractional position in the horizontal and vertical directions from an integer grid of reconstructed reference samples. Typically, a two-dimensional interpolation is performed on the reconstructed reference grid values of the integer samples to obtain the sample values at the location of the fractional sample offset. The process of obtaining predicted samples from reconstructed reference pictures using a pair of candidate MVs can be carried out in one of the following ways:

Round the fractional part of a pair of original MVs to the nearest integer location and obtain the integer grid values of the reconstructed reference pictures.

Perform 2-tap (e.g., bilinear) separable bilinear interpolation to obtain predicted sample values to the fractional pixel precision specified by the source MV pair.

Perform divisive interpolation with a higher tap (e.g., 8 or 6 taps) to obtain predicted sample values to the fractional pixel precision specified by the source MV pair.

Although candidate MV pairs can have an arbitrary subpixel offset relative to the source MV pair, in some embodiments, for ease of searching, candidate MV pairs are selected with an integer pixel distance relative to the source MV pair. In such cases, predicted samples over the entire pair of candidate MVs can be obtained by performing prediction on a block of samples around the pair of original MVs to cover all refinement positions around the pair of original MVs.

In some embodiments, once the dissimilarity cost value in all pairs of candidate MVs at an integer distance from the pair of original MVs has been estimated, additional candidate MV pairs are added at sub-pixel distance offsets from the position of the best cost value. Predicted samples are obtained for each of these positions using one of the methods described earlier, and the cost of dissimilarity is estimated and compared to obtain the position of least dissimilarity. In some other embodiments, to avoid this computationally expensive process of predicting for each subpixel distance position around the best-cost integer distance position, the estimated integer distance cost values are stored and the parametric error surface is set to a neighborhood of the best integer distance. The minimum of this error surface is then analytically calculated and used as the position with minimum dissimilarity. In such cases, the dissimilarity cost value is said to be derived from the calculated integer distance cost values.

The application of motion vector refinement for a given encoded sample block may be dictated by certain encoding properties of the encoded sample block. Some examples of such encoding properties could be:

the distance in number of pictures (when sampling at the same frame rate) from the current picture to the two reference pictures used for bidirectional prediction of the coded block of samples is the same and falls on opposite sides of the current picture.

the initial dissimilarity between two predicted blocks obtained using a pair of original MVs is less than a predefined threshold for each sample.

Bidirectional Predictive Optical Flow (BPOF) refinement

Bidirectional predictive optical flow refinement is the process of improving the bidirectional prediction accuracy of a block without explicit additional signaling in the bit stream other than that specified for bidirectional prediction. This is part of the inter prediction block (244) in FIG. 2 and 344 in Fig. 3.

In bidirectional prediction, 2 outer predictions are obtained according to two motion vectors, then the predictions are combined by applying weighted averaging. Combined prediction can result in reduced residual energy as quantization noise in the two reference patches is cancelled, thereby providing greater encoding efficiency than unidirectional (single) prediction. Weighted combination in bidirectional prediction can be done using the equation:

Bi-prediction = Prediction1 * W1 + Prediction2 * W2 + K,

where W1 and W2 are weights that may be signaled in the bitstream or may be predetermined on the encoder side or on the decoder side. K is an additive coefficient, which can also be signaled in the bitstream or predetermined on the encoder side or on the decoder side. For example, bidirectional prediction can be obtained using

Bi-prediction = (Prediction1 + Prediction2)/2,

where W1 and W2 are set to 1/2 and K is set to 0.

The purpose of optical flow refinement is to improve the accuracy of bidirectional prediction. Optical flow is the pattern of apparent movement of objects in an image between two successive frames. Optical flow is caused by the movement of an object or camera. The optical flow refinement process improves the accuracy of bidirectional prediction by applying the optical flow equation (solving the optical flow equation).

In the example, pixel I (x, y, t) is located in the first frame (x and y correspond to the spatial coordinates, t correspond to the time dimension). The object represented by the pixel moves a distance (dx, dy) in the next frame taken after time dt. Since these pixels are the same and the intensity does not change, the optical flow equation is defined as follows:

I (x, y, t) = = I (x+dx , y+dy , t+dt)

I(x, y, t) indicates the intensity (sampling value) of a pixel at coordinates (x, y, t).

In another example, small displacements and higher order terms in the Taylor series expansion are ignored, the optical flow equations can also be written as:

= 0

Where And are the horizontal and vertical gradients of the spatial sample at position (x, y), and - partial derivative with respect to time at point (x, y).

Optical flow refinement uses the above principle to improve the quality of bidirectional prediction.

Implementation of optical flow refinement usually includes the following steps:

1. Calculation of sampling gradients;

2. Calculate the difference between the first prediction and the second prediction;

3. Calculate the offset of pixels or group of pixels that minimizes the error between two reference patches obtained using the optical flow equation

where I⁽⁰⁾corresponds to the sample value in the first prediction, I⁽¹⁾ is the sample value in the second prediction, and ∂I⁽⁰⁾/ ∂x and ∂I⁽⁰⁾/ ∂y - gradients in -x and -y directions.Anddenote the distances to the reference pictures on which the first prediction and the second prediction are obtained. Motion vector (v_x,v_y) is obtained by the process minimization. Some approaches minimize the sum of squared errors, while some approaches minimize the sum of absolute errors.

4. Using an implementation of the optical flow equation such as below:

Where predBIO points to a modified prediction that is the result of an optical flow refinement process.

Approximate gradients can be obtained using the following formula

∂I (x, y, t) / ∂x=I (x+1, y, t) - - I (x - 1, y, t)

∂I (x, y, t) / ∂y=I (x, y+1, t) - - I (x, y - 1, t)

In some embodiments, to simplify the complexity of estimating the offset for each pixel, the offset is estimated for a group of pixels. In some examples, to calculate improved bidirectional prediction for a 4x4 luma sample block, the offsets are estimated using the sample values of the 8x8 luma sample block with the 4x4 luma sample block at its center.

The input to the optical flow refinement process is the prediction samples from two reference pictures, and the output of the optical flow refinement is a combined prediction (predBIO), which is calculated according to the optical flow equation.

One example of optical flow refinement is explained in section 8.4.7.4 “Bidirectional optical flow prediction process” of JVET-M1001, Versatile Video Coding (Draft 4).

The terms “optical flow refinement,” “bidirectional predictive optical flow refinement,” and “bidirectional optical flow refinement” are used interchangeably in the disclosure because the terms are equivalent in nature.

In the example, motion vector refinement and optical flow refinement are applied sequentially as follows:

Step 0, obtain the initial motion vectors as in block 1010 of FIG. 8.

In step 1, the motion vector refinement block 1020 is applied, and the refined vectors of the motion vector block 1030 are obtained.

In step 2, predictions are obtained according to the motion vectors of the refinement block 1040. The obtained predictions are I ⁽⁰⁾ and I ⁽¹⁾ , which are inputs to the optical flow refinement process.

Step 3, the optical flow refinement process is applied to the predictions to obtain a modified prediction. The modified prediction is obtained according to the optical low-level equation and is denoted pred _BIO .

However, the optical flow refinement process requires large computational resources. Decoding time is increased by applying optical flow refinement.

In one embodiment of the present invention, a method for deciding whether to apply optical flow refinement or not is disclosed, which decision can be made according to calculations performed during the motion vector refinement process.

More specifically, the result of calculations performed during the motion vector refinement process is used to determine whether to apply optical flow refinement or not.

The purpose of the invention is to skip the application of optical flow refinement according to a given condition in order to reduce the average decoding time (by skipping the necessary calculations).

According to the first exemplary embodiment, the following steps are applied to obtain a prediction for the current encoding block:

Step 0: Obtain the initial motion vectors based on the indication information in the bit stream.

Step 1: Obtain first predictions based on the original motion vectors and the M-tap interpolation filter.

Step 2: obtain comparable value according to first predictions.

Step 3: Obtain the refined motion vector according to the original motion vectors and matching cost.

Step 4: Obtain the second prediction according to the refined motion vector and K-tap interpolation filter.

Step 5: Determine whether to perform the optical flow refinement process according to the matching cost. In one example, the matching cost is compared to a threshold, and the optical flow refinement process is performed when the matching cost value is greater than or equal to the threshold. Step 5 can also be performed before Step 3 or Step 4.

Step 6: When it is determined that the optical flow refinement process needs to be performed, optical flow refinement is applied with the second prediction as input and the modified second prediction as output. If defined negatively, the optical flow refinement is not applied to the second prediction. In other words, when it is determined that the optical flow refinement process needs to be performed, the final prediction of the current encoding block is obtained according to the second prediction and according to the optical flow refinement process. Otherwise, the final prediction of the current encoding block is obtained according to the second prediction and without applying the optical flow refinement process.

The detailed explanation of the steps is as follows:

In step 0, two initial motion vectors are received as input. The initial motion vectors may be determined based on the indication information in the bitstream. For example, an index may be transmitted in a bitstream, the index indicating a position in a list of candidate motion vectors. In another example, the motion vector predictor index and the motion vector difference value may be signaled in the bitstream. The motion vector, which is determined based on the indication information in the bit stream, is determined as the original motion vectors.

In another example, the reference picture indications may be obtained from the bit stream, the original motion vectors are obtained based on the reference picture indications. Reference picture indications are used to determine the reference pictures that the source motion vectors point to.

Stage 1, stage 2 and stage 3 correspond to the motion vector refinement process as explained in the examples above. The initial motion vectors are refined according to the motion vector refinement. In one example, the matching cost is a measure of similarity that is used in motion vector refinement.

According to step 1, the first predictions corresponding to the original motion vectors are obtained. In the example, there are at least 2 pairs of candidate motion vectors in the motion vector refinement process, one of which is usually the pair formed by the original motion vectors (MV0, MV1). In other words, the set of candidate motion vectors typically includes more than one pair, with one of the pairs usually (MV0, MV1). Another pair of candidate motion vectors is determined based on (MV0, MV1) by adding small perturbations to the motion vectors (as explained in the examples above).

In step 1, the first predictions corresponding to each pair of candidate motion vectors are obtained based on the M-tap interpolation filter. For example, one prediction corresponding to MV0 can be obtained by placing a rectangular block in a reference picture (a picture that is already encoded in the encoder or decoded in the decoder), with the block indicated by MV0. An interpolation filter is then preferentially applied to the samples in the block indicated by MV0. To provide more accurate motion estimation, the resolution of the reference picture can be improved by interpolating samples between pixels. Fractional pixel interpolation can be performed by weighted averaging of nearby pixels. Here, an M-tap filter can typically be a 2-, 4-, 6-, or 8-tap filter (but not limited to these parameters), which means that the filter has M multipliers. The prediction corresponding to MV1 can be obtained in a similar way by placing a rectangular block in the same or a different reference picture. The size of the rectangle is proportional to the size of the current encoding block.

In step 2, the matching cost associated with each pair of candidate motion vectors is determined according to the first predictions.

According to step 2, at least one matching cost (eg, a similarity measure) corresponding to one of the motion vector (MV) pairs of the refinement candidate is obtained. The higher the similarity between two prediction blocks, the lower the matching cost.

The specified matching cost is used when refining the original motion vectors in step 3. The refined motion vector is selected according to the mentioned matching cost.

In step 4, the second prediction is obtained according to the refined motion vector and the K-tap interpolation filter. In the case of two refined motion vectors (MV0' and MV1'), which is the case of bidirectional prediction, two second predictions are obtained.

The second prediction is obtained by applying a second interpolation filter (K-tap filter), which may or may not be identical to the first interpolation filter (M-tap filter). The second prediction is obtained similarly to the first prediction using the second interpolation filter and in accordance with the block specified by MV0' and MV1' in the reference picture.

In step 5, said matching cost is used to determine whether to perform the optical flow refinement process or not, according to the following:

when the matching cost value is less than a predetermined threshold, optical flow refinement is not applied. When the matching cost value is greater than or equal to the threshold value, the optical flow refinement process is performed. If the optical flow refinement process is performed, the final prediction samples are changed.

In step 6, according to the output of step 5, if the matching cost is greater than r equal to the mentioned threshold, the optical flow refinement process is applied to the second predictions, the second predictions are obtained according to MV0' and MV1' (refined motion vectors). The final prediction for the current encoding block is obtained by performing an optical flow refinement process on the second predictions, the second predictions being indicated by MV0' and MV1'. If the matching cost is less than the specified threshold, the final prediction is obtained according to the second predictions specified by MV0' and MV1' without applying optical flow refinement, which means that step 6 is not performed.

In one implementation, the matching cost in step 2 is the matching cost corresponding to a pair of original motion vectors (which is one of the candidate refinement motion vector (MV) pairs). The matching cost can correspond to the pair MV0, MV1.

In another implementation, said matching cost in step 2 is a matching cost equal to the smallest matching cost among pairs of candidate motion vectors (MVs) for refinement. In other words, a matching cost corresponding to each pair of candidate motion vectors for refinement is obtained, and said matching cost is equal to the smallest matching cost among them. In one example, said matching cost is the matching cost corresponding to the refined motion vector pair MV0' and MV1' (since the refined motion vector pair (MV0', MV1') is selected because it has the lowest matching cost.

For example, MV pairs can be constructed as follows.

Pairs of candidate MVs are determined by adding small motion vector differences to MV0 and MV1. For example, candidate MV pairs could include the following:

(MV0, MV1)

(MV0 + (0,1), MV1 + (0, -1))

(MV0 + (1.0), MV1 + (-1.0))

MV0 and MV1 are the original motion vectors, MV0’ and MV1’ are the refined motion vectors throughout the application.

According to another implementation, when the optical flow refinement process is not performed, the final prediction is obtained according to the following formula:

Bi-prediction = Prediction1 * W1 + Prediction2 * W2 + K,

where W1 and W2 are weights, W1 and W2 may be transmitted in the bit stream, or W1 and W2 may be predetermined at the encoder side or at the decoder side. K is an additive coefficient, which can also be signaled in the bitstream or predetermined on the encoder side or on the decoder side. In an example, bidirectional prediction can be obtained using

Bi-prediction = (Prediction1 + Prediction2)/2,

where W1 and W2 are set to 1/2 and K is set to 0. Prediction1 and prediction2 are the second predictions that are obtained using K-tap interpolation filtering, Prediction1 corresponds to the first refined MV (MV0'), and Prediction2 corresponds to the second refined MV ( MV1').

The above equation produces a weighted combination of the two predictions, and the result is the final prediction for the block.

The threshold may be a predefined value, the value of the threshold may depend on the size of the prediction block. For example, the threshold could be th=nCbW x nCbH x K, where K is a value greater than zero, nCbW and nCbH are the width and height of the prediction block.

The first embodiment is further illustrated by a block diagram in FIG. 6.

In one implementation, an M-tap filter is a 2-tap filter (eg, a bilinear filter) with one of the taps set to zero. In this implementation, the M-tap filter uses 2 multipliers, one of which is always zero. Which coefficient has a value equal to zero is determined based on the fractional sampling point when the fractional sampling point is indicated by the motion vector. In this case, the value of the first multiplication factor or the value of the second multiplication factor may be zero, depending on the fractional component of the motion vector.

An example of such a filter (with two taps, one of which is zero) can be illustrated in accordance with the following table:

Fractional sample position p interpolation filter coefficients fb _L [p][0] fb _L [p][1] 1 16 0 2 16 0 3 16 0 4 16 0 5 16 0 6 16 0 7 16 0 8 0 16 9 0 16 10 0 16 eleven 0 16 12 0 16 13 0 16 14 0 16 15 0 16

The fractional sampling position (p) can be obtained according to the components of the original or refined motion vector. For example, if the -x component of the motion vector is specified as MV0x, then the fractional sampling position can be obtained as p=MV0x% 16, where "%" is the modulo operation. Typically, p=MV0x% K, where K represents the number of fractional sample positions between two sample positions. The interpolation filter shown as an example above may also be called a single-tap filter because only one of the filter's taps is non-zero at any given time.

In one implementation, the value of K is 8. In other examples, the value of M is less than 8.

In one implementation, the value of M and the value of K are 8.

The original motion vector is obtained in block 710, which is the input of the motion vector refinement block. The search space is built around the original motion vector by the motion vector refinement block (740). In one example, the search space consists of pairs of candidate motion vectors, the first motion vector of the pair corresponding to the first reference picture, and the second motion vector of the pair corresponding to the second reference picture. The first predictions corresponding to each pair of candidate motion vectors are obtained at step 710 by applying an M-tap interpolation filter. As part of the motion vector refinement, a matching cost corresponding to one of the pairs of motion vectors in the search space (720) is calculated. Said matching cost is used as part of two processes, the first process is motion vector refinement (740), where the matching cost is used to decide which motion vector pair is selected as the refined motion vector pair (750). The second process is deciding whether optical flow refinement (770) is applied or not. After receiving the refined motion vector, a second prediction for the current block is obtained by block (760). If the matching cost is greater than or equal to the threshold, optical flow refinement is applied and the prediction at step 760 is modified at step 770 to obtain a modified prediction (780). The modified prediction typically differs in sample values from the second prediction at step 760.

In one example, the motion vector refinement process is performed more than once to further refine the motion vector. In this example, the original motion vectors are first refined by a motion vector refinement process to obtain the first refined motion vector. After this, the motion vector refinement is performed again, in this case the first refined motion vector is considered as the original motion vectors for the second motion vector refinement.

According to a second exemplary embodiment, the following steps are applied to obtain a prediction for the current encoding block:

Step 0: Obtain the initial motion vectors based on the indication information in the bit stream.

Step 1: Obtain first predictions based on the original motion vectors and the M-tap interpolation filter.

Stage 2: obtain N comparable values according to the first predictions.

Step 3: Obtain the refined motion vector according to the original motion vectors and N matching costs based on the first function.

Step 4: Obtain the second prediction according to the refined motion vector and K-tap interpolation filter.

Step 5: Determine whether to perform the optical flow refinement process according to N matching costs. The inferred value is obtained according to N comparable values and the second function. In the example, the inferred cost is compared with a threshold value, and the optical flow refinement process is performed when the inferred cost value is greater than or equal to the threshold value. Step 5 can also be performed before Step 3 or Step 4.

Step 6: Modify at least one prediction sample of the current encoding block using optical flow refinement when it is determined that the optical flow refinement process needs to be performed.

When it is determined that the optical flow refinement process needs to be performed, optical flow refinement is applied with the second prediction as input and the modified second prediction as output. If defined negatively, the optical flow refinement is not applied to the second prediction. In other words, when it is determined that the optical flow refinement process needs to be performed, the final prediction of the current encoding block is obtained according to the second prediction and according to the optical flow refinement process. Otherwise, the final prediction of the current encoding block is obtained according to the second prediction and without applying the optical flow refinement process.

The detailed explanation of the steps is as follows:

In step 0, two initial motion vectors are received as input. The initial motion vector can be determined based on the indication information in the bitstream. For example, an index may be signaled in a bitstream, the index indicating a position in a list of candidate motion vectors. In another example, the motion vector predictor index and the motion vector difference value may be signaled in the bitstream. The motion vector, which is determined based on the indication information in the bit stream, is determined as the original motion vectors.

In another example, reference picture indications may be obtained from a bit stream, the original motion vectors are obtained based on the reference picture indications. Reference picture indications are used to determine the reference pictures that the source motion vectors point to.

Stage 1, stage 2 and stage 3 correspond to the motion vector refinement process as explained in the examples above. The initial motion vectors are refined according to the motion vector refinement. In one example, the matching cost is a measure of similarity that is used in motion vector refinement.

According to step 1, the first predictions corresponding to the original motion vectors are obtained. In the example, there are at least 2 pairs of candidate motion vectors in the motion vector refinement process, one of which is usually the pair formed by the original motion vectors (MV0, MV1). And another pair of candidate motion vectors is determined based on (MV0, MV1) by adding small perturbations to the motion vectors (as explained in the above example).

In step 1, the first predictions corresponding to each pair of candidate motion vectors are obtained based on the M-tap interpolation filter.

In step 2, N matching costs associated with N pairs of candidate motion vectors are determined according to the first predictions.

According to step 2, N matching costs (similarity measure) corresponding to N pairs of candidate motion vectors (MVs) for refinement are obtained. The higher the similarity between two prediction blocks, the lower the matching cost.

The specified N comparison costs are used when refining the initial motion vectors at stage 3.

The refined motion vector is determined according to the first function and N matching costs.

In one example, the refined motion vector may be obtained in accordance with the following function:

if (sad[3]+sad[5]) is equal to (sad[4]<<1), dmvOffset[0] is set to 0,

Otherwise the following applies:

dmvOffset[0]=((sad[3]-sad[5])<<3)/(sad[3]+sad[5]-(sad[4]<< 1))

if (sad[1]+sad[7]) is equal to (sad[4]<<1), dmvOffset[1] is set to 0,

Otherwise the following applies:

dmvOffset[1] = ((sad[1]-sad[7])<<3)/(sad[1]+sad[7]-(sad[4]<<1))

where dmvOffset [0] and dmvOffset [1] define the difference between the original and refined motion vectors. In the example, dmvOffset[0] and dmvOffset[1] define the -x and -y components of the difference between the refined and original motion vectors. sad[0] to sad[7] are N matching costs corresponding to N pairs of candidate motion vectors. The refined motion vector is obtained by adding dmvOffset to the original motion vectors.

There may be other functions that can be used to determine the refined motion vector according to N matching costs. The first function in the invention is not limited to the above equation.

In step 4, the second prediction is obtained according to the refined motion vector and the K-tap interpolation filter. In the case of two refined motion vectors (MV0' and MV1'), which is the case of bidirectional prediction, two second predictions are obtained.

The second prediction is obtained by applying a second interpolation filter (K-tap filter), which may or may not be identical to the first interpolation filter (M-tap filter). The second prediction is obtained similarly to the first prediction using the second interpolation filter and in accordance with the block specified by MV0' and MV1' in the reference picture.

In step 5, the inferred cost is obtained according to the second function and the mentioned N agreed costs. The inferred cost is used to determine whether to perform the optical flow refinement process or not. When the value of said derived cost is less than a predetermined threshold, the optical flow refinement process is not performed. When the value of the inferred cost is greater than or equal to the threshold value, the optical flow refinement process is performed. If the optical flow refinement process is performed, the final prediction samples are modified.

In step 6, according to the output of step 5, if the inferred cost is greater than the specified threshold, the optical flow refinement process is applied to the second predictions, the second predictions are obtained according to MV0' and MV1' (refined motion vectors). The final prediction for the current encoding block is obtained by performing an optical flow refinement process on the second predictions, the second predictions being indicated by MV0' and MV1'. If the matching cost is less than the specified threshold, the final prediction is obtained according to the second predictions specified by MV0' and MV1' without applying optical flow refinement, which means that step 6 is not performed.

According to another implementation, when the optical flow refinement process is not performed, the final prediction is obtained according to the following formula:

Bi-prediction = Prediction1 * W1 + Prediction2 * W2 + K,

where W1 and W2 are weights, W1 and W2 may be transmitted in the bit stream or may be predetermined at the encoder side or decoder side. K is an additive coefficient, which can also be signaled in the bitstream or predetermined on the encoder side or on the decoder side. In an example, bidirectional prediction can be obtained using

Bi-prediction = (Prediction1 + Prediction2)/2 ,

where W1 and W2 are set to 1/2 and K is set to 0. Prediction1 and prediction2 are the second predictions that are obtained using K-tap interpolation filtering, Prediction1 corresponds to the first refined MV (MV0'), and Prediction2 corresponds to the second refined MV (MV1').

The above equation produces a weighted combination of the two predictions, and the result is the final prediction for the block.

The threshold may be a predefined value, the value of the threshold depends on the size of the prediction block. For example, the threshold could be th=nCbW x nCbH x K, where K is a value greater than zero, nCbW and nCbH are the width and height of the prediction block.

The second embodiment is further illustrated by a block diagram in FIG. 7.

In one implementation, an M-tap filter is a 2-tap filter (eg, a bilinear filter) with one of the taps set to zero. In this implementation, the M-tap filter uses 2 multipliers, one of which is always zero. The coefficient equal to zero is determined based on the fractional sampling point pointed to by the motion vector. In this case, the value of the first multiplication factor or the value of the second multiplication factor may be zero, depending on the fractional component of the motion vector.

An example of such a filter (with two taps, one of which is zero) can be illustrated in accordance with the following table:

Fractional sample position p interpolation filter coefficients fb _L [p][0] fb _L [p][1] 1 16 0 2 16 0 3 16 0 4 16 0 5 16 0 6 16 0 7 16 0 8 0 16 9 0 16 10 0 16 eleven 0 16 12 0 16 13 0 16 14 0 16 15 0 16

The fractional sample position (p) can be obtained according to the components of the original or refined motion vector. For example, if the -x component of the motion vector is given as MV0x, then the fractional sampling position can be obtained as p=MV0x% 16, where "%" is the modulo operation. Typically, p=MV0x% K, where K represents the number of fractional sample positions between two sample positions. The interpolation filter shown as an example above may also be called a single-tap filter because only one of the filter's taps is non-zero at any given time.

Another example of a bilinear interpolation filter could be the following, in which case both filter coefficients are non-zero:

Fractional sample position p interpolation filter coefficients fb _L [p][0] fb _L [p][1] 1 15 1 2 14 2 3 13 3 4 12 4 5 eleven 5 6 10 6 7 9 7 8 8 8 9 7 9 10 6 10 eleven 5 eleven 12 4 12 13 3 13 14 2 14 15 1 15

In one implementation, the value of K is 8. In other examples, the value of M is less than 8.

In one implementation, the value of M and the value of K are 8.

In one implementation, the second function may be a function for linearly combining the N matching costs according to dmvOffset, where dmvOffset was obtained in step 3. The linear combination of x and y will be any expression of the form ax+by, where a and b are constants. In the example, the constants a and b can be defined based on dmvOffset. Examples for the second function are given below.

In one implementation, the second function could be:

Sad [1] * A+Sad [2] * B+Sad [3] * C+Sad [4] * D, where A, B, C and D are greater than or equal to zero. In one example, A, B, C and D can be numbers from 0 to 1, and add up to 1 (i.e. A+B+C+D=1). In another example, A, B, C and D may be numbers greater than or equal to 0 and which add up to a predetermined fixed number P, P may be 1,2, 4, 8, 16, etc.

A, B, C and D can be predefined fixed numbers.

A, B, C and D can be obtained according to dmvOffset[0] and dmvOffset[1]. In the example, A=dmvOffset [0], B=P1 - dmvOffset [0], C=dmvOffset [1], D=P2 - dmvOffset [1]. Where P1 and P2 can be 1, 4, 8,16, etc.

The above equation is given as an example. The equation is a linear combination of the 4 comparable values to obtain the inferred value. The equation uses dmvOffset, which can be obtained in step 3. dmvOffset represents the difference between the refined motion vector and the original motion vector. In one particular implementation, dmvOffset is defined as the difference between MV0 and MV0'. In particular, dmvOffset[0] can be the difference between the -x component for MV0 and MV0', while dmvOffset[1] can be the difference between the -y component for MV0 and MV0'.

In another implementation, the second function could be:

Sad [1] * A+Sad [2] * B+Sad [3] * C, where A, B and C are greater than or equal to zero. In one example, A, B and C can be numbers from 0 to 1, and add up to 1 (i.e. A+B+C=1). In another example, A, B and C may be numbers greater than or equal to 0 and which add up to a predetermined fixed number P, P may be 1, 2, 4, 8, 16, etc.

A, B and C can be predefined fixed numbers.

A, B and C can be obtained according to dmvOffset[0] and dmvOffset[1]. In the example, A=P - dmvOffset [0] - dmvOffset [1], B=dmvOffset [0], C=dmvOffset [1]. Where P can be 1, 4, 8, 16, etc.

The above equation is given as an example. The equation is a linear combination of the three comparable values to obtain the inferred value. The equation uses dmvOffset, which can be obtained in step 3. dmvOffset represents the difference between the refined motion vector and the original motion vector. In one example, dmvOffset is defined as the difference between MV0 and MV0'. In particular, dmvOffset[0] can be the difference between the -x component for MV0 and MV0', while dmvOffset[1] can be the difference between the -y component for MV0 and MV0'.

In another implementation, the second function to get the inferred value could be:

Using 5 estimated dissimilarity cost values (e.g., SAD values) in the refined MV pair and candidate MV pairs that are an integer distance from the refined MV pair, a parametric error surface function

E(x,y) = A*(x - x ₀ ) ² + B*(y - y ₀ ) ² + C

Selected where₍x0_,y0) corresponds position at which the dissimilarity between two reference patches is minimized, C is the cost value at (x0, y0), and A, B - coefficients models. These 5 unknowns can be solved exactly if 5 cost values are available. In other words, the equation for E(x,y) makes the assumption that the shape of the matching costs as a function of spatial positions around the position with the minimum matching costs has a parabolic shape.

In one embodiment, pairs of candidate MVs to the left, top, right, and bottom of the refined MV pair are used at a distance of one whole pixel. In this case, given the estimated values of E(x, y) at (x, y) positions (0,0), (-1,0), (0, -1), (1,0) and (0,1 ) and parametric equation for E (x, y), 5 unknowns A, B, C_,x0, y0 can be solved as follows:

A = (E (-1.0) + E (1.0) - 2 * E (0.0)) / 2

B = (E (0, -1) + E (0.1) - 2 * E (0.0)) / 2

x0 = _{(E (-} 1.0) - E (1.0)) / (2 * (E (-1.0) + E (1.0) - 2 * E (0.0)))

y0 = _(E (0, -1) - E (0,1)) / (2 * (E (0, -1) + E (0,1) - 2 * E (0,0)))

C=

On the other hand, if cost values in more than 5 positions are available, the 5 unknowns can be solved using least squares or a similar approach. The resulting value of C becomes the inferred cost.

In one implementation, the second function could be:

Where K is a scalar greater than 0, and sad[0] to sad[4] are N matching costs.

In one example, the motion vector refinement process is performed more than once to further refine the motion vector. In this example, the original motion vectors are first refined by a motion vector refinement process to obtain the first refined motion vector. After this, the motion vector refinement is performed again, in this case the first refined motion vector is considered as the original motion vectors for the second motion vector refinement.

The original motion vector is obtained at step 925, which is the input of the motion vector refiner block. The search space is built around the original motion vector by the motion vector refinement block (930). In one example, the search space consists of pairs of candidate motion vectors, the first motion vector of the pair corresponding to the first reference picture, and the second motion vector of the pair corresponding to the second reference picture. The first predictions corresponding to each pair of candidate motion vectors are obtained at step 910 by applying an M-tap interpolation filter. As part of the motion vector refinement, a matching cost corresponding to the N pairs of motion vectors in the search space (915) is calculated. The N matching costs are used as part of two processes, the first process is motion vector refinement (930), where the matching costs are used to compute a refined pair of motion vectors (935) according to a function that takes the N matching costs as input. The second process is the decision of whether optical flow refinement (950) is applied or not, where the decision is made by step 945. After receiving the refined motion vector, a second prediction for the current block is obtained by (940). If the matching cost exceeds the threshold, optical flow refinement is applied and the prediction at step 940 is modified at step 950 to obtain a modified prediction (955-960). The modified prediction typically differs in sample values from the second prediction at step 940. If the matching cost is less than a threshold, no optical flow refinement is applied and the second prediction is set as the output prediction (the final prediction of the current block).

According to a third exemplary embodiment of the present invention, the following steps are applied to obtain a prediction for the current encoding block:

Step 0: Obtain a pair of initial motion vectors based on the indication information in the bitstream.

Step 1: Obtain the first set of predicted samples based on the pair of original MVs and the M-tap interpolation filter.

Step 2: Obtain the first matching cost corresponding to a pair of original MVs using the first set of predicted samples.

Step 3: Determine whether the current encoding block is suitable for performing motion vector refinement.

Step 4: If the current encoding block is determined to be eligible to perform MVR in step 3,

Step 4a: Obtain the refined MV pair and matching cost corresponding to the refined MV pair according to the original MV pair and matching cost using the motion vector refinement process.

Step 4b: Obtain the second set of predicted samples according to the refined MV pair and the K-tap interpolation filter.

Step 4c: Determine whether to perform the optical flow refinement process according to the second matching cost. In one example, the matching cost is compared to a threshold, and the optical flow refinement process is performed when the matching cost value is greater than or equal to the threshold.

Step 5: Otherwise (if the current encoding block is determined to be ineligible for MVR execution in step 3),

Step 5a: Obtain a second set of predicted samples according to a pair of original MVs and a K-tap interpolation filter.

Step 5b: Determine whether to perform the optical flow refinement process according to the first matching cost. In one example, the matching cost is compared to a threshold, and the optical flow refinement process is performed when the matching cost value is greater than or equal to the threshold.

Step 6: When it is determined that the optical flow refinement process needs to be performed (either in step 4c or step 5b), optical flow refinement is applied with the second prediction as input and the modified second prediction as output. If defined negatively, the optical flow refinement is not applied to the second prediction. In other words, when it is determined that the optical flow refinement process needs to be performed, the final prediction of the current encoding block is obtained according to the second prediction and according to the optical flow refinement process. Otherwise, the final prediction of the current encoding block is obtained according to the second prediction and without applying the optical flow refinement process.

This embodiment is further illustrated in the block diagram of FIG. 9. Block 1110 receives a pair of source MVs for the current prediction encoding block at positions L0 and L1. Block 1110 corresponds to step 1, in which a first set of predicted samples is obtained using a pair of original MVs and reconstructed reference picture samples L0 and L1. Block 1120 corresponds to step 2, in which a first matching cost (or dissimilarity metric such as SAD) is estimated between the first set of predicted sample blocks corresponding to a pair of original MVs (as described in the background MVRs section). Block 1130 corresponds to step 3, which checks whether the current encoding block matches the MVR execution. Block 1140 corresponds to step 4a where, if the current encoding block is determined to be eligible to perform MVR, a refined MV pair is obtained by performing the MVR (as described in the background MVR section) and a second matching cost (or dissimilarity metric) corresponding to the refined MV pair . Block 1150 corresponds to step 4b, in which a second set of predicted samples is obtained using a K-tap interpolation filter (in the horizontal and vertical directions) using the refined MV pair. Block 1160 corresponds to step 4c, which checks whether the second matching cost is less than a predetermined threshold below which bidirectional predictive optical flow refinement and bidirectional prediction are skipped. Block 1180 corresponds to step 5a, where the current encoding block skips the MVR and obtains a second set of predicted samples using a K-tap interpolation filter using the pair of original MVs. Block 1185 corresponds to step 5b, which checks whether the cost of the first match is less than a predetermined threshold below which the BPOF is passed. Blocks 1170 and 1195 correspond to a portion of step 6 in which, if the check in step 4c or step 5b indicates that the second or first matching cost, respectively, is less than a predetermined threshold below which BPOF is passed, averaging the weighted bidirectional prediction without BPOF is performed using the second set predicted samples. Block 1175 corresponds to a portion of step 6 in which, if the check in step 4c or step 5b indicates that the second or first matching cost is not less than a predetermined threshold below which BPOF is passed, an optical flow estimate is obtained and a final bidirectional prediction is obtained using the second set predicted samples, gradients of the second set of predicted samples, and estimated optical flow.

It should be noted that when using the first or second matching cost, which are calculated for a sub-block of samples in the current encoding block, as required by the motion vector refinement process to determine early completion of the bidirectional predictive optical flow-based refinement process, the decision to skip or perform BPOF may vary from one MVR subblock to another within an encoding block. BPOF shall be applied or omitted for all BPOF application blocks (eg, pixel level or 4x4 sample block level) within a sub-block based on the determination made in step -4c or step -5.

In some embodiments, it is possible to perform additional early termination for each BPOF application block in a subblock by obtaining partial comparable costs corresponding to each BPOF application block in an MVR subblock.

The predetermined threshold value is typically selected as a per-sample threshold that depends on the bit depth of the first prediction or the first set of predicted samples. For example, if the first prediction sample value obtained using bilinear (2-tap) interpolation is limited by the bit depth b, the threshold value for each sample is calculated as k * 2 ^{(b - 10)} , and the number of samples for which the matching cost is calculated is is N, then the predefined threshold against which the matching cost for the current subblock is compared must be ^{k *} N * 2 ^(10-b) . Example values for k are 2 (for bit width 10), N is 8×16=128, and b is 8. Since the matching cost for a given pair of candidate MVs can be calculated with a thinned set of the first predicted samples, the value of N should be used accordingly. For example, if interleaved rows of an 8x16 block of predicted samples are used, N should be calculated as 8x8=64.

According to embodiments of the invention, an early termination method is provided to conditionally skip the application of the optical flow refinement process, the application of the optical flow refinement process is considered computationally intensive. As a result, the average decoding time is reduced.

In addition, a condition for conditionally skipping the application of optical flow is determined based on parameters that are calculated by another process (calculating matching costs during the motion vector refinement process). Since already calculated values are used, no additional calculations are necessary.

Specifically, a video encoding method implemented in a decoding apparatus or an encoding apparatus is provided, as shown in Figure 10. The method comprises the following steps, which may be performed in a predetermined order. The initial motion vector is obtained for the current block 1210. The current block may be the current encoding block. The first predictions for the sample value in the current block are obtained based on the original motion vectors 1220. The matching cost is calculated according to the first 1230 predictions.

After obtaining the first matching cost, it is determined whether the optical flow refinement process 1240 is to be performed or not according to at least one predetermined condition, wherein the at least one predefined condition comprises a condition of whether the calculated matching cost is (e.g., in terms of similarity measures; see description above) equal to or greater than a predetermined threshold value. The optical flow refinement process to obtain the final inter-prediction for the sample value in the current block is performed at step 1250 when it is determined that the optical flow refinement process is to be performed. When it is determined that the optical flow refinement process does not need to be performed, computational costs can be reduced by skipping the optical flow refinement process.

This method can be implemented in the devices described above with reference to Figures 1a to 5.

In particular, the method may be implemented in the context of a motion vector refinement process on the decoder side. The input data for such a process is:

luminance location (xSb, ySb), defining the top left sample of the current encoding subblock relative to the top left luminance sample of the current picture,

the sbWidth variable, which determines the width of the current encoding subblock in brightness samples,

the sbHeight variable, which determines the height of the current encoding subblock in brightness samples,

brightness motion vectors with an accuracy of 1/16 fractional sampling mvL0 and mvL1,

selected arrays of samples of reference brightness pictures refPicL0L and refPicL1L.

The outputs of this process are the delta brightness motion vectors dMvL0 and dMvL and the variable dmvrSad, which defines the minimum sum of the absolute differences of the first predictions (see the SAD calculation described above).

The delta luminance motion vector dMvL0 can be inferred using dMvL0[0]+=16*intOffX and dMvL0[1]+=16*intOffY, where intOffX and intOffY are integer sample offsets in the x and y directions, respectively. Additionally, the delta luminance motion vector dMvL can be calculated as dMvL1[0] = -dMvL0[0] and dMvL1[1] = -dMvL0[1].

The values of the first predicted brightness sample are derived using fractional bilinear interpolation of the sample. The inter-prediction block decoding process may or may not apply a bidirectional optical flow sample prediction process. If not applied, the weighted sample prediction process is applied to the refined second predictions obtained from the refined motion vectors. If a bidirectional optical flow sample prediction process is applied, it takes the second predictions obtained from the refined motion vectors as input and outputs the final predictions.

The flag may be used to signal whether the bidirectional optical flow sample prediction process can be applied. For example, it may be considered a necessary condition for the execution of the bidirectional optical flow sample prediction process that the flag be TRUE. However, this necessary condition may not be a sufficient condition to perform the bidirectional optical flow sample prediction process. A sufficient condition may be that both the flag is TRUE and the matching cost described above is equal to or greater than a predefined threshold. For example, the cost of matching can be determined based on the variable dmvrSad, which defines the minimum sum of the absolute differences of the first predictions.

On the other hand, it may be considered a sufficient condition to not perform the optical flow sample prediction process but to perform the weighted sample prediction process if the flag is FALSE.

Additionally, an apparatus 1300 is provided for use in an image encoder and/or an image decoder, as shown in FIG. 11. The device 1300, according to this exemplary embodiment, includes a source motion vector block 1310, which is configured to obtain source motion vectors for the current block. In addition, device 1300 includes a prediction block 1320 that is configured to obtain first predictions for the sample value in the current block based on the original motion vectors. In addition, the device 1300 includes a matching cost calculation unit 1330, which is configured to calculate the matching cost according to the first predictions.

The apparatus 1300 includes an optical flow refinement process determination unit 1340, which is configured to determine whether to perform an optical flow refinement process according to at least one preset condition, wherein the at least one predefined condition comprises a condition of whether the calculated matching cost is equals or exceeds the threshold value. In addition, the device 1300 includes an optical flow refinement process execution unit 1350, which is configured to perform an optical flow refinement process to obtain a final inter-prediction for the sample value in the current block when it is determined that the optical flow refinement process is to be performed.

Mathematical operators

The mathematical operators used in this application are similar to those used in the C programming language. However, the results of integer division and arithmetic shift operations are defined more precisely, and additional operations such as exponentiation and real-valued division are also defined. Numbering and counting conventions specify starting at 0, for example, "first" is equivalent to 0, "second" is equivalent to 1, etc.

Arithmetic operators

The following arithmetic operators are defined as follows:

+ Addition - Subtraction (as a two-argument operator) or negation (as a unitary prefix operator) * Multiplication, including matrix multiplication x ^y Exponentiation. Defines x to the power of y. In other contexts, this notation is used as a superscript that should not be interpreted as exponentiation. / Integer division with the result truncated towards zero. For example, 7/4 and -7/-4 are truncated to 1, and -7/4 and 7/-4 are truncated to -1. ÷ Used to indicate division in mathematical equations where truncation or rounding is not intended. Used to indicate division in mathematical equations where truncation or rounding is not intended. Summation f(i), where i takes all integer values from x to y inclusive. x % y Modulo operation. The remainder of x divided by y is defined only for integers x and y with x >= 0 and y > 0.

Logical operators

The following logical operators are defined as follows:

x && y Boolean logical "and" for x and y

x || y Boolean logical "or" for x and y

! Boolean logical "not"

x? y:z If x is TRUE or not 0, the value of y is calculated; otherwise, the z value is calculated.

Relational operators

The following relational operators are defined as follows:

> More

>= Greater than or equal to

< Less

<= Less than or equal to

= = Equal

!= Not equal

When a relational operator is applied to a syntax element or variable that is assigned the value "na" (not applicable), the value "na" is treated as a separate value for the syntax element or variable. The value "na" is not considered equal to any other value.

Bit operators

The following bitwise operators are defined as follows:

& Bitwise "and". When working with integer arguments, operates on the two's complement representation of an integer value. When working on a binary argument that contains fewer bits than another argument, the shorter argument is expanded by adding more significant bits that are 0.

| Bitwise "or". When working with integer arguments, operates on the two's complement representation of an integer value. When working on a binary argument that contains fewer bits than another argument, the shorter argument is expanded by adding more significant bits that are 0.

^ Bitwise "or only". When working with integer arguments, operates on the two's complement representation of an integer value. When working on a binary argument that contains fewer bits than another argument, the shorter argument is expanded by adding more significant bits that are 0.

x >> y Arithmetic right shift of the two's complement integer representation of x by y binary digits. This function is defined only for non-negative integer values of y. Bits shifted to the most significant bits (MSB) by a right shift have a value equal to MSB x before the shift operation.

x << y Arithmetic left shift of the two's complement integer representation of x by y binary digits. This function is defined only for non-negative integer values of y. Bits shifted to the least significant bits (LSB) by left shifting have a value of 0.

Assignment Operators

The following arithmetic operators are defined as follows:

= assignment operator

+ + Increment, i.e. x+ + is equivalent to x = x +1; when used in an array index, evaluates the value of the variable before the increment operation.

- - Decrease, i.e. x- - is equivalent to x = x- 1; when used in an array index, evaluates the value of the variable before the decrement operation.

+= Increment by the specified amount, i.e. x += 3 is equivalent to x=x+3, and x += (-3) is equivalent to x=x + (-3).

-= Decrease by the specified amount, i.e. x -= 3 is equivalent to x=x - 3, and x -= (-3) is equivalent to x=x - (-3).

Range designation

The following notation is used to indicate a range of values:

x=y..z x takes integer values from y to z, inclusive, where x, y and z are integers and z is greater than y.

Mathematical functions

The following mathematical functions are defined:

Abs( x ) =

Asin(x) is a trigonometric inverse sine function operating on an argument x that ranges from -1.0 to 1.0 inclusive, with an output value ranging from -π ÷ 2 to π ÷ 2 inclusive in radians

Atan(x) is a trigonometric arctangent function operating on an argument x, with an output value ranging from -π ÷ 2 to π ÷ 2 inclusive, in radians.

Atan2(y, x) =

(where English otherwise – otherwise)

Ceil( x ) smallest integer greater than or equal to x

Clip1Y( x ) = Clip3( 0, ( 1 << BitDepthY ) - 1, x )

Clip1C( x ) = Clip3( 0, ( 1 << BitDepthC ) - 1, x )

Clip3( x, y, z ) =

Cos( x ) is a trigonometric cosine function that operates on x in radians.

Floor( x ) is the largest integer less than or equal to x.

GetCurrMsb( a, b, c, d ) =

Ln( x ) natural logarithm of x (logarithm to base e, where e is the base constant of natural logarithm 2.718281828…).

Log2( x ) logarithm of x to base 2

Log10( x ) is the logarithm of x to base 10.

Min( x, y ) =

Max( x, y ) =

Round( x ) = Sign( x ) * Floor( Abs( x ) + 0.5)

Sign( x ) =

Sin( x ) is a trigonometric sine function that takes x in radians.

Sqrt( x ) =

Swap( x, y ) = ( y, x )

Tan( x ) is a trigonometric tangent function that takes the argument x in radians.

Order of priority of operations

When the order of precedence in an expression is not explicitly specified using parentheses, the following rules apply:

-- Higher priority operations are evaluated before any lower priority operation.

-- Operations with the same priority are evaluated sequentially from left to right.

The table below lists the priority of operations from highest to lowest; a higher position in the table indicates higher priority.

For those operators that are also used in the C programming language, the order of precedence used in this description is the same as in the C programming language.

Table: Operation priority from highest (at the top of the table) to lowest (at the bottom of the table)

operations (with operands x, y and z) "x++", "x- -" "!x", "-x" (as unary prefix operator) xy "x * y", "x/y", "x ÷ y", " %y" "x+y", "x - y" (as an operator with two arguments), " " "x << y", "x >> y" "x < y", "x <= y", "x > y", "x >= y" "x == y", "x != y" "x&y" "x | y" "x && y" "x | | y" "x ? y : z" "x..y" "x=y", "x += y", "x -= y"

Text description of logical operations

The text contains a definition of logical operations, which would be described mathematically in the following form:

if( condition 0 )

statement 0

else if( condition 1 )

statement 1

...

else /* informative remark on remaining condition */
statement n

can be described as follows:

... as below / ... the following applies:

- if condition 0, definition 0

- Otherwise, if condition 1, definition 1

-...

-- Otherwise (informative comment on the remaining condition), definition n

Each "if...otherwise, if...else,..." definition in the text is introduced by the words "...as follows" or "...the following applies" followed immediately by "if...". The last condition “if ... otherwise, if ... otherwise, ...” is always “otherwise, ...”. Alternating definitions "if...else, if...else,..." can be identified by matching "...as follows" or "...the following applies" with the final "otherwise,...".

The text contains a definition of logical operations, which would be described mathematically in the following form:

if( condition 0a && condition 0b )

statement 0

else if( condition 1a || condition 1b )

statement 1

...

else

statement n

can be described as follows:

... as below / ... the following applies:

-- If all of the following conditions are true, definition 0:

- condition 0a

- condition 0b

-- Else, if one or more of the following conditions are true, definition 1:

- condition 1a

- condition 1b

-...

- otherwise, definition n

The text contains a definition of logical operations, which would be described mathematically in the following form:

if( condition 0 )

statement 0

if( condition 1 )

statement 1

can be described as follows:

When condition is 0, definition is 0

When condition 1, definition 1

Although embodiments of the invention have been primarily described in terms of video encoding, it should be noted that embodiments of encoding system 10, encoder 20, and decoder 30 (and thus system 10) and other embodiments described herein may also be configured for processing or encoding a still picture, that is, processing or encoding a single picture independently of any previous or subsequent picture, as in video encoding. In general, only inter prediction (encoder) and 344 (decoder) blocks 244 may not be available if the picture processing encoding is limited to a single picture 17. All other functionality (also referred to as tools or technologies) of video encoder 20 and video decoder 30 may be equally degree be used for still picture processing, such as residual computation 204/304, transform 206, quantization 208, inverse quantization 210/310, (inverse) transform 212/312, division 262/362, intra prediction 254/354, and/or loop filtering 220, 320 and entropy encoding 270 and entropy decoding 304.

Embodiments of, for example, encoder 20 and decoder 30, as well as the functions described herein, for example, with respect to encoder 20 and decoder 30 may be implemented by hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on a computer-readable medium or transmitted through a communication medium in the form of one or more instructions or code and executed by a hardware processor. Computer-readable media may include computer-readable media that corresponds to a tangible medium, such as a storage medium, or a communication medium, including any medium that enables transmission of a computer program from one place to another, such as according to a communication protocol. Thus, computer readable media may typically correspond to (1) tangible computer readable media that is durable, or (2) a communication medium such as a signal or carrier wave. Storage media can be any available media that can be accessed by one or more computers or one or more processors to retrieve instructions, code and/or data structures for implementing the techniques described in this disclosure. The computer program product may include computer-readable media.

By way of example, and not limitation, such computer-readable media may comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, flash memory, or any other media that can used to store desired program code in the form of instructions or data structures and can be accessed by a computer. Additionally, any connection is properly referred to as a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave communications, then coaxial cable, fiber optic cable, twisted pair, DSL or wireless technologies such as infrared, radio and microwave communications are included in the definition of media/media. However, it should be understood that computer readable storage media and storage media do not include connections, carrier waves, signals or other short-lived media, but are instead directed to long-term, tangible storage media. The disc and disc used in this document include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc, where discs typically reproduce data magnetically, and the disks reproduce the data optically using lasers. Combinations of the above should also be included in the scope of computer readable media.

The instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, application-specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other equivalent integrated or discrete logic circuits. Accordingly, the term “processor” as used herein can refer to any of the above structure or any other structure suitable for implementing the techniques described herein. Additionally, in some aspects, the functionality described herein may be provided within specialized hardware and/or software modules configured for encoding and decoding or included in a bundled codec. In addition, the techniques may be fully implemented in one or more circuits or logic elements.

The techniques of this disclosure may be implemented in a wide variety of devices or apparatus, including a wireless telephone, an integrated circuit (IC), or a set of ICs (eg, a chipset). This disclosure describes various components, modules, or blocks to highlight functional aspects of devices configured to implement the disclosed technologies, but not necessarily required to be implemented by different hardware units. Rather, as described above, the various blocks may be combined into a codec hardware block or provided by a collection of interacting hardware blocks, including one or more processors as described above, together with suitable software and/or firmware.

Claims (31)

1. A video encoding method implemented in a decoding device or encoding device, containing: obtaining initial motion vectors for the current block; obtaining first predictions for the sample value in the current block based on the original motion vectors; calculating a first matching cost according to the first predictions; determining whether the optical flow refinement process should be performed or not according to at least one predetermined condition, wherein the at least one predetermined condition comprises a condition of whether the calculated first matching cost is equal to or greater than a certain threshold; And performing an optical flow refinement process to obtain a final inter-prediction for the sample value in the current block when it is determined that the optical flow refinement process is to be performed; wherein the method additionally contains: obtaining refined motion vectors based on the original motion vectors and the first matching cost; obtaining second predictions for the sample value in the current block according to the refined motion vectors; And performing an optical flow refinement process comprising performing an optical flow refinement based on the second predictions. 2. The method of claim 1, wherein the at least one predefined condition comprises a condition that the current block can be predicted by refining a motion vector at the decoder side. 3. The method of claim 1 or 2, wherein it is determined that the optical flow refinement process is to be performed when it is determined that all of the at least one predetermined condition are satisfied. 4. The method as claimed in one of the preceding claims, wherein the first predictions for the sample value in the current block are obtained based on the first interpolation filter. 5. The method of claim 4, wherein the first interpolation filter is a bilinear interpolation filter. 6. Method according to any one of paragraphs. 1-5, wherein obtaining first predictions for a sample value in the current block based on the original motion vectors comprises: obtaining a number of candidate pairs based on the original motion vectors; obtaining first predictions for the sample value in the current block based on at least one of the candidate pairs; and wherein calculating a first matching cost according to the first predictions comprises determining a matching cost for each of the candidate pairs based on the first predictions and determining the smallest matching cost from the determined matching costs as the first matching cost. 7. The method of claim 6, wherein second predictions for the sample value in the current block are obtained according to the second interpolation filter. 8. The method of claim 6 or 7, wherein the second interpolation filter is a 6- or 8-tap interpolation filter. 9. Method according to one of paragraphs. 6-8, characterized in that the refined motion vectors are obtained according to the second matching cost. 10. The method of claim 9, wherein when the value of the second matching cost is greater than or equal to another threshold value, it is determined that the optical flow refinement process should be performed. 11. Method according to one of paragraphs. 6-10, in which, only when it is determined that the optical flow refinement process should not be performed, the final external prediction is obtained by a weighted sum of the second predictions. 12. The method according to one of the preceding claims, wherein the threshold value is obtained in accordance with the size of the current block. 13. The method according to one of the preceding claims, wherein the current block is an encoding block or a subblock. 14. The method of one of the preceding claims, further comprising generating an inter prediction block containing a final inter prediction for the sample value in the current block. 15. An encoder (20) containing a processing circuit for performing the method according to any one of paragraphs 1-14. 16. A decoder (30) comprising a processing circuit for performing the method according to any one of claims. 1-14. 17. A non-volatile computer-readable storage medium containing program code for executing the method according to any one of claims. 1-14.

Images