CN114286090A

CN114286090A - Intra-frame prediction method and device for image

Info

Publication number: CN114286090A
Application number: CN202011035054.3A
Authority: CN
Inventors: 虞露; 李道文
Original assignee: Zhejiang University ZJU
Current assignee: Zhejiang University ZJU
Priority date: 2020-09-27
Filing date: 2020-09-27
Publication date: 2022-04-05

Abstract

The invention provides an intra prediction method and device for an image. Obtaining affine deformation information of a current block from a bit stream, obtaining affine deformation vectors of at least two control points of the current block by using the affine deformation information, and obtaining affine deformation vectors of at least two sub-blocks in the current block, which are different, by using an affine model, the affine deformation vectors of the control points of the current block and the positions of the control points of the current block; obtaining a prediction sample of at least one sub-block according to the affine deformation vector of the sub-block and a sample in the current image; in obtaining the sub-block prediction samples, if the samples in the current picture include samples required for prediction but not yet decoded, the not yet decoded samples are generated from decoded samples adjacent to the not yet decoded sample position in the current picture. The method utilizes the spatial affine relation between the non-adjacent areas of the image, increases the utilization rate of the reference sample of the undecoded area, improves the intra-frame prediction accuracy and improves the coding efficiency.

Description

Intra-frame prediction method and device for image

Technical Field

The present invention relates to video image processing technologies, and in particular, to a method and an apparatus for intra prediction of an image.

Background

Video image compression is used in many current and emerging products such as digital television set-top boxes, high definition television decoders, digital versatile disc players and personal computers, and the like. Without video image compression, the digital video content may be so large that it is difficult or even impossible to efficiently store, transmit, and view the digital video content. Over the years, various video coding and decoding methods for compressing digital video content have been studied and various video image compression standards have been developed to standardize video image coding methods so that compressed digital video content can be stored and transmitted in formats recognizable by most video decoders. For example, the Moving Picture Experts Group (MPEG) and the international telecommunication union (ITU-T) have developed widely used video image codec standards.

Most video image coding standards use various coding techniques such as temporal and spatial prediction, transform and quantization, entropy coding to achieve data compression. Two types of prediction techniques, intra-prediction and inter-prediction, are commonly included in video codecs to improve compression efficiency. Intra-frame prediction exploits the spatial correlation of pixels within a video frame or image, while inter-frame prediction exploits the temporal correlation between video frames.

In intra prediction, the reference samples for the samples to be predicted are from the already decoded region of the video frame currently being decoded. Intra-prediction techniques include local prediction techniques and non-local prediction techniques. The local prediction technology uses a decoded sample adjacent to the spatial position of the sample to be predicted to perform extrapolation, so as to obtain a predicted value of the sample to be predicted. Non-local prediction technology is also called intra block copy (intra block copy), and in a decoded area of a video frame currently being decoded, a reference sample of a sample to be predicted is found by using a translation matching method, wherein a spatial displacement of the sample to be predicted relative to the reference sample is called a block vector (block vector).

Inter-frame prediction techniques involve two important concepts, reference frames and motion vectors. The reference frame refers to a video frame where a reference sample of a sample to be predicted is located, and for the inter-frame prediction technology, the reference frame is other decoded video frames which are not currently being decoded. A motion vector refers to the cross-temporal motion displacement of the sample to be predicted relative to the reference sample. Most video codec standards consider object motion in a picture to conform to rigid motion and follow a translational motion model. However, the real world movement is diverse, and irregular movements such as zooming and rotating are ubiquitous. Now, video coding experts have recognized the popularity of irregular motion, and hope to improve the inter-frame prediction accuracy by introducing an irregular motion model (e.g., affine motion model), thereby improving the coding efficiency.

For a four-parameter affine motion model, different motion vectors of two control points in a region to be predicted need to be obtained, and then the motion vector of a sample to be predicted is derived according to the motion vector of the control point, the position of the control point and the position of a certain sample to be predicted and the four-parameter affine motion model. As shown in FIG. 1A, two control points of the region to be predicted are respectively the upper left pixel of the region to be predicted and the upper right pixel of the region to be predicted, the positions of the control points are respectively (0,0) and (W-1,0), and the motion vectors of the control points are respectively (0, W-1,0)

And

wherein the content of the first and second substances,

displaced by horizontal movement

And vertical movement displacement

Composition of, is

Displaced by horizontal movement

And vertical movement displacement

Composition of, is

If the position of a certain sample to be predicted is (x, y), the affine motion vector of the sample to be predicted

The following model was used:

for a six-parameter affine motion model, different motion vectors of three control points in a region to be predicted need to be obtained, and then the motion vector of a sample to be predicted is derived according to the motion vector of the control point, the position of the control point and the position of a certain sample to be predicted and according to the six-parameter affine motion model. As shown in FIG. 1B, the three control points of the region to be predicted are the upper left pixel of the region to be predicted, the upper right pixel of the region to be predicted and the lower left pixel of the region to be predicted, the positions of the control points are (0,0), (W-1,0) and (0, W-1), the motion vectors of the control points are (0,0), (W-1,0) and (W-1), respectively

And

wherein the content of the first and second substances,

displaced by horizontal movement

And vertical movement displacement

Composition of, is

Displaced by horizontal movement

And vertical movement displacement

Composition of, is

Displaced by horizontal movement

And vertical movement displacement

Composition of, is

If the position of a sample to be predicted is (x, y), the affine motion vector { mv of the sample to be predicted^x,mv^yThe following model gave:

although inter-prediction techniques have introduced various prediction techniques based on affine motion models, intra-prediction techniques are still in the exploration phase. In fact, even within the same video frame, there are sometimes non-linear relationships such as scaling, rotation, etc. between different regional pictures. For example, parts such as a handrail and a window are repeatedly arranged on a building in the actual three-dimensional world, and even if the actual sizes of the parts are the same, due to the perspective imaging principle, the zooming phenomenon occurs in the shooting content of the building; if the imaging surface and the building surface are not strictly parallel, the building photographic content can also have a rotation phenomenon. Also for example, in recent years, short videos are getting hot, and users often add some special effects with spatial scaling or spatial rotation relationships. Under the occasions, the intra-frame prediction technology considering the complex linear relation can effectively improve the prediction accuracy, thereby improving the compression efficiency. At present, an expert tries to use a spatial affine transformation model, after obtaining a preliminary reference sample by using an intra-frame block matching method, performs affine transformation on the preliminary reference sample, and uses the affine transformation result as a final reference sample. However, the processing flow is complex, and affine transformation model parameters can only be selected from a small number of sets or even fixed singly, so that the application universality and effectiveness of the scheme are limited. Some other experts try to use similar inter-frame affine prediction technology for reference, perform spatial affine deformation on block vectors of control points in intra-frame non-local prediction technology, thereby deriving block vectors of other prediction points in the region to be predicted, and perform spatial deformation compensation by using the derived block vectors, to obtain pixels at positions pointed by the block vectors as reference samples, as shown in fig. 2A and 2B. Here, a block vector based on a spatial affine deformation is named as an affine deformation vector. However, the technique based on deriving the affine deformation vector uses the inter-frame prediction technique for reference, and thus, when performing spatial deformation compensation, only the reference samples can be obtained from the decoded area of the image being decoded. This results in reference samples that are sufficiently valuable in some extreme cases to be considered invalid, such as reference samples located in an undecoded region adjacent to a decoded region of the image being decoded, thereby affecting prediction accuracy and compression efficiency.

Accordingly, there is a need in the art for a method and apparatus for intra prediction of an image being decoded that overcomes the above-mentioned deficiencies.

Disclosure of Invention

In order to overcome the defects of the intra-frame prediction technology, the invention provides an intra-frame prediction technology and a device for an image being decoded, which utilize the spatial affine relation between a region to be predicted and a non-adjacent region thereof in the image being decoded, and increase the utilization rate of reference samples in an undecoded region, thereby improving the intra-frame prediction accuracy and further improving the coding efficiency.

The design idea of the invention is as follows: and utilizing a spatial affine relation between the to-be-predicted area and a non-adjacent reference area thereof in the image being decoded, using corresponding spatial deformation vectors of at least two control points in the to-be-predicted area in the non-adjacent reference area, deriving spatial deformation vectors of the to-be-predicted samples except the control points in the to-be-predicted area according to an affine deformation model, and obtaining predicted values of the to-be-predicted samples from the non-adjacent reference area according to the derived spatial deformation vectors of the to-be-predicted samples. When obtaining the predicted value of the sample to be predicted, if the reference sample in the un-decoded area needs to be used, the reference sample value in the un-decoded area is filled by the adjacent decoded sample.

The invention proposes a method for intra prediction of an image being decoded, comprising:

obtaining affine deformation information of a current block from a bit stream, obtaining affine deformation vectors of at least two control points of the current block by using the affine deformation information, and obtaining affine deformation vectors of at least two sub-blocks in the current block, which are different, by using an affine model, the affine deformation vectors of the control points of the current block and the positions of the control points of the current block;

obtaining a prediction sample of at least one sub-block according to the affine deformation vector of the sub-block and a sample in the current image; in obtaining the sub-block prediction samples, if the samples in the current picture include samples required for prediction but not yet decoded, the not yet decoded samples are generated from decoded samples adjacent to the not yet decoded sample position in the current picture.

Meanwhile, the present invention proposes an intra prediction apparatus for an image being decoded, which includes:

affine deformation information analysis module: parsing and outputting affine deformation information of the current block from an input bitstream,

a control point affine deformation vector calculation module: calculating and outputting affine deformation vectors of at least two control points of the current block by using the input affine deformation information of the current block,

a sub-block affine deformation vector calculation module: calculating and outputting affine deformation vectors which are different and are of at least two sub-blocks in the current block from the input affine deformation vector of the control point of the current block and the position of the control point of the current block by using an affine model,

a predicted sample acquisition module: deriving and outputting prediction samples of at least one sub-block from the input affine deformation vector of the sub-block and samples in the current image,

the prediction sample acquisition module comprises an undecoded sample generation module and a prediction sample derivation module,

the undecoded sample generation module: if the samples in the current image comprise samples which are needed by prediction but not decoded yet, inputting decoded samples which comprise positions adjacent to the samples which are not decoded yet in the current image, generating and outputting the samples which are not decoded yet according to a generating rule,

the prediction sample derivation module: deriving and outputting prediction samples for at least one sub-block from the input affine deformation vector of the sub-block, decoded samples in a current image and the generated not-yet-decoded samples.

Compared with the prior art, the invention has the beneficial effects that:

according to the method, the spatial affine relation between the to-be-predicted area and the non-adjacent reference area is utilized in the image being decoded, the spatial deformation vectors corresponding to at least two control points in the to-be-predicted area in the non-adjacent reference area are used, the spatial deformation vectors of the to-be-predicted samples except the control points in the to-be-predicted area are deduced according to the affine deformation model, the predicted values of the to-be-predicted samples are obtained from the non-adjacent reference area according to the deduced spatial deformation vectors of the to-be-predicted samples, the situation that affine transformation model parameters can be selected from a small number of sets and are even fixed singly is avoided, the affine transformation model parameters can be obtained in a self-adaptive mode according to the content of each area of the image, and meanwhile, the utilization rate of the reference samples in the non-decoded area is improved, and therefore the intra-frame prediction accuracy is further improved, and the encoding efficiency is improved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.

FIG. 1A is a diagram of an inter-frame prediction four-parameter affine motion model.

FIG. 1B is a diagram of a six-parameter affine motion model for inter-frame prediction.

Fig. 2A is a schematic diagram of an intra-frame prediction four-parameter affine deformation model.

Fig. 2B is a diagram illustrating an intra-frame prediction six-parameter affine deformation model.

Fig. 3 is a basic block diagram of a decoder of a conventional video codec standard.

Fig. 4 is a flowchart of an intra prediction apparatus for an image according to an embodiment of the present invention.

Fig. 5 is a schematic diagram of neighboring blocks used for inheritance-based DVP list construction.

Fig. 6 is a schematic diagram of neighboring blocks used in construction based DVP lists.

Fig. 7 is a schematic diagram of an intra prediction apparatus for an image according to an embodiment of the present invention.

Fig. 8 is a schematic diagram of an intra prediction apparatus for an image according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Some concepts that may be involved with embodiments of the present invention are described below. In summary, a video image decoder is shown in fig. 2. In the decoder, the entropy decoding unit 123 is used to restore syntax in the bitstream. The reconstructed transform coefficients obtained after decoding and parsing are subjected to an Inverse Quantization (IQ) process and an Inverse Transform (IT) process (IQ + IT,116) to reconstruct the prediction residual. While the decoded syntax element values are processed through the intra/inter sub-blocks to output intra/inter prediction data to the reconstructor 117. The reconstructed prediction residual is then added back to the intra/inter prediction data in reconstructor 117 to reconstruct the image data. Here, the reconstructed image data is a reconstructed pixel value, and the operation of the reconstructed pixel value is referred to as prediction compensation processing. To reduce noise in the reconstructed pixel values, the filtering process is performed using one or more filtering techniques, such as a deblocking filter 119 and a sample adaptive offset filter 121. The reconstructed pixel values output after the filtering process are stored in

buffers

118, 120, and 122, or the filtering process is continued using a filtering technique for decoding of subsequent images.

The present specification describes an intra prediction method and apparatus that can be used for both intra-type frames and inter-type frames in a video image codec.

Example 1

The present example provides a method for intra prediction of an image, which specifically includes:

obtaining affine deformation vectors of two control points of a current block from a bit stream, deriving affine deformation vectors of two sub-blocks in the current block and different affine deformation vectors by using a four-parameter affine model, affine deformation vectors of the two control points of the current block and spatial positions of the two control points, and performing prediction compensation in a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. When prediction compensation is carried out, samples in a current image comprise samples which are needed by prediction but not decoded yet, and the average value of all decoded samples is calculated according to all decoded samples in the same row with the not decoded samples in the current image, and is used as the not decoded samples.

The spatial accuracy of the affine deformation vectors of the control points and the sub-blocks involved in the process is integer pixel accuracy, and the spatial accuracy of the affine deformation vectors of the control points and the sub-blocks used in all the areas when the current image is decoded is integer pixel accuracy. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 2

obtaining affine deformation vectors of two control points of a current block from a bit stream, deriving affine deformation vectors of two sub-blocks in the current block and different affine deformation vectors by using a four-parameter affine model, affine deformation vectors of the two control points of the current block and spatial positions of the two control points, and performing prediction compensation in a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. When prediction compensation is carried out, samples in a current image comprise samples which are needed by prediction but not decoded yet, and the average value of the decoded samples is calculated according to at least two decoded samples in the current image which are in the same column with the samples which are not decoded yet, and the average value is used as the samples which are not decoded yet.

Example 3

obtaining affine deformation vectors of two control points of a current block from a bit stream, deriving affine deformation vectors of two sub-blocks in the current block and different affine deformation vectors by using a four-parameter affine model, affine deformation vectors of the two control points of the current block and spatial positions of the two control points, and performing prediction compensation in a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. When prediction compensation is carried out, samples in a current image comprise samples which are needed by prediction but are not decoded yet, according to all decoded samples in the current image, which are in the same row and the same column with the samples which are not decoded yet, an intra-frame direction prediction method is used, all decoded samples are used for calculating generated samples according to a 45-degree prediction direction, and the generated samples are used as the samples which are not decoded yet.

Example 4

obtaining affine deformation vectors of two control points of a current block from a bit stream, deriving affine deformation vectors of two sub-blocks in the current block and different affine deformation vectors by using a four-parameter affine model, affine deformation vectors of the two control points of the current block and spatial positions of the two control points, and performing prediction compensation in a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. When prediction compensation is carried out, samples in a current image comprise samples which are needed by prediction but are not decoded, searching is carried out towards the right (upper, lower) direction by taking the leftmost sample position of the image in the same row with the sample not decoded as a search starting point, the first decoded sample is selected as the generated sample, if the search range of the right (upper, lower) direction exceeds the image range and the decoded sample is not searched yet, searching is carried out towards the lower (left, right) direction by taking the uppermost sample position of the image in the same row with the sample not decoded as a search starting point, and the first decoded sample is selected as the generated sample. The generated sample is taken as the not yet decoded sample.

Example 5

obtaining affine deformation vectors of two control points of a current block from a bit stream, deriving affine deformation vectors of two sub-blocks in the current block and different affine deformation vectors by using a four-parameter affine model, affine deformation vectors of the two control points of the current block and spatial positions of the two control points, and performing prediction compensation in a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. When prediction compensation is carried out, samples in a current image comprise samples which are needed by prediction but are not decoded, a search is carried out in the left direction by taking the position of the nearest sample on the left side of the sample which is not decoded as a search starting point, the first decoded sample is selected as the generated sample, if the search range in the left direction exceeds the image range and the decoded sample is not searched, the position of the nearest sample on the upper side (the right side and the lower side) of the sample which is not decoded is taken as the search starting point, a search is carried out in the upper side (the right side and the lower side) direction, and the first decoded sample is selected as the generated sample. The generated sample is taken as the not yet decoded sample.

Example 6

obtaining affine deformation vectors of two control points of a current block from a bit stream, deriving affine deformation vectors of two sub-blocks in the current block and different affine deformation vectors by using a four-parameter affine model, affine deformation vectors of the two control points of the current block and spatial positions of the two control points, and performing prediction compensation in a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. When prediction compensation is carried out, samples in a current image comprise samples which are needed by prediction but are not decoded, the top nearest sample position of the sample which is not decoded is used as a search starting point, search is carried out in the top direction, the first decoded sample is selected as the generated sample, if the search range in the top direction exceeds the image range, the decoded sample is not searched, the left (right, lower) nearest sample position of the sample which is not decoded is used as the search starting point, search is carried out in the left (right, lower) direction, and the first decoded sample is selected as the generated sample. The generated sample is taken as the not yet decoded sample.

Example 7

obtaining affine deformation vector predictions of three control points of a current block from a bit stream, performing equivalent processing on the affine deformation vector predictions of the three control points, taking the affine deformation vector predictions of the three control points as affine deformation vectors of the three control points, using a six-parameter affine model, the affine deformation vectors of the three control points of the current block and the spatial positions of the three control points, deriving two sub-block affine deformation vectors which are different in the current block, and performing prediction compensation on a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial accuracy of the control point affine deformation vector prediction, the control point affine deformation vector and the affine deformation vector of the sub-block involved in the process is integer pixel accuracy, and the spatial accuracy of the affine deformation vector of the control point and the affine deformation vector of the sub-block used in all the areas when the current image is decoded is integer pixel accuracy. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 8

obtaining affine deformation vector predictions of three control points of a current block and affine deformation vector differences among the three control points and the three affine deformation vector predictions from a bit stream, adding the affine deformation vector predictions of the three control points and the affine deformation vector differences of the three control points to obtain affine deformation vectors of the three control points, using a six-parameter affine model, the affine deformation vectors of the three control points of the current block and spatial positions of the three control points to derive affine deformation vectors of two sub-blocks in the current block, and performing prediction compensation in a current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial accuracy of the control point affine deformation vector prediction, the control point affine deformation vector and the affine deformation vector of the sub-block involved in the process is integer pixel accuracy, and the spatial accuracy of the affine deformation vector of the control point and the affine deformation vector of the sub-block used in all the areas when the current image is decoded is integer pixel accuracy. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 9

obtaining an index value CPDVP idx of an affine deformation vector predictor CPDVP (control point displacement vector predictor) of two control points of the current block from the bit stream. Constructing an inheritance-based subblock affine deformation prediction reference DVP (displacement vector prediction) list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₁To A₀The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₁Using A₁Control point affine deformation vector of₁The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₁To B₀To B₂If the first block satisfying the condition is B₁Using B₁Control point affine formVariable vector, B₁And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

the set constitutes a third CPDVP construction candidate. And predicting a CPDVP index value CPDVP idx according to the affine deformation vectors of the two control points of the current block, and finding the affine deformation vector prediction CPDVP of the two control points of the current block corresponding to the CPDVP idx in the inherited DVP-based list and the constructed DVP-based list. Performing equivalence processing on the affine deformation vector predictions CPDVPs of the two control points, taking the affine deformation vector predictions DPDVP of the two control points as affine deformation vectors CPDV (control point displacement vector) of the two control points, deriving affine deformation vectors of two sub-blocks in the current block and different from each other by using a four-parameter affine model, the affine deformation vectors CPDV of the two control points of the current block and the spatial positions of the two control points, and performing prediction compensation in the current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the process is integer pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the sub-block used in all the areas is integer pixel precision when the current image is decoded. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 10

and obtaining the affine deformation vector prediction CPDVP index value CPDVP idx of the two control points of the current block from the bit stream. Build an inheritance-based DVP list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₀Using A₀Control point affine deformation vector of₀The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₀Using B₀Control point affine deformation vector, B₀And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). Dividing affine deformation vectors of neighboring blocks of the current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group,the CPDVP candidates of the first control point CP1 of the current block are derived from the second group, and the CPDVP candidates of the second control point CP2 of the current block are derived from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

the set constitutes a third CPDVP construction candidate. And predicting a CPDVP index value CPDVP idx according to the affine deformation vectors of the two control points of the current block, and finding the affine deformation vector prediction CPDVP of the two control points of the current block corresponding to the CPDVP idx in the inherited DVP-based list and the constructed DVP-based list. Performing equivalent processing on the affine deformation vector predictions CPDVP of the two control points, and taking the affine deformation vector predictions DPDPDPDPDPVP of the two control points as the affine deformation of the two control pointsAnd the variable vector CPDV derives the affine deformation vectors of the two sub-blocks in the current block, which are different from each other, by using a four-parameter affine model, the affine deformation vectors CPDV of the two control points of the current block and the space positions of the two control points, and carries out prediction compensation on the current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the process is integer pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the sub-block used in all the areas is integer pixel precision when the current image is decoded. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 11

and obtaining affine deformation vector prediction CPDVP index values CPDVP idx of the three control points of the current block from the bit stream. Build an inheritance-based DVP list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (2) checks neighboring blocks in the left block group, and if there are no blocks that satisfy the condition, takes the CPDVP formed by zero values as the first affine CPDVP inheritance candidate. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂The search order of (2) checking neighboring blocks in the upper block groupAnd if no block meets the condition, taking the CPDVP formed by the zero values as a second affine CPDVP inheritance candidate. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

collectionConstitute a first CPDVP construction candidate. And predicting a CPDVP index value CPDVP idx according to the affine deformation vectors of the three control points of the current block, and finding the affine deformation vector prediction CPDVP of the three control points of the current block corresponding to the CPDVP idx in the inherited DVP-based list and the constructed DVP-based list. Performing equivalence processing on affine deformation vector predictions CPDVP of the three control points, taking the affine deformation vector predictions DPDVP of the three control points as affine deformation vectors CPDV of the three control points, using a six-parameter affine model, the affine deformation vectors CPDV of the three control points of the current block and the spatial positions of the three control points, deriving affine deformation vectors which are different and are of two sub-blocks in the current block, and performing prediction compensation on the current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the process is integer pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the sub-block used in all the areas is integer pixel precision when the current image is decoded. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 12

and obtaining the affine deformation vector prediction CPDVP index value CPDVP idx of the two control points of the current block from the bit stream. Build an inheritance-based DVP list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed by the above-mentioned components,according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₀Using A₀Control point affine deformation vector of₀The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₀Using B₀Control point affine deformation vector, B₀And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises block vectors of neighboring blocks A

Block vector of neighboring block B

Block vector of neighboring block C

The second group comprises block vectors of neighboring blocks D

And block vectors of neighboring blocks E

The third group comprises block vectors of neighboring blocks F

And block vectors of neighboring blocks G

Checking in the search order A through C through B if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 0. Checking in the search order E to D if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 1. Checking in the search order G through F if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 2.

The set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

the set constitutes a third CPDVP construction candidate. And predicting a CPDVP index value CPDVP idx according to the affine deformation vectors of the two control points of the current block, and finding the affine deformation vector prediction CPDVP of the two control points of the current block corresponding to the CPDVP idx in the inherited DVP-based list and the constructed DVP-based list. Performing equivalence processing on affine deformation vector predictions CPDVPs of the two control points, taking the affine deformation vector predictions DPDVPs of the two control points as affine deformation vectors CPDVs of the two control points, using a four-parameter affine model, the affine deformation vectors CPDVs of the two control points of the current block and the space positions of the two control points, deriving affine deformation vectors of the two sub-blocks in the current block, performing prediction compensation on the current image according to the affine deformation vectors of the two sub-blocks, and obtaining prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the process is integer pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the sub-block used in all the areas is integer pixel precision when the current image is decoded. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 13

and obtaining the affine deformation vector prediction CPDVP index value CPDVP idx of the two control points of the current block from the bit stream. Build an inheritance-based DVP list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₀Using A₀Control point affine deformation vector of₀The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₀Using B₀Control point affine deformation vector, B₀And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises block vectors of neighboring blocks A

Block vector of neighboring block B

Block vector of neighboring block C

The second group comprises block vectors of neighboring blocks D

And block vectors of neighboring blocks E

The third group comprises block vectors of neighboring blocks F

And block vectors of neighboring blocks G

Checking in the search order A through B through C if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 0. Checking in the search order D to E if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 1. Checking in the search order from F to G if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 2.

The set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

Example 14

obtaining affine deformation vector prediction CPDVP index values CPDVP idx of three control points of the current block and affine deformation vector difference values CPDVD (control point displacement vector difference) between the affine deformation vectors CPDV of the three control points and the three affine deformation vector prediction CPDVP from the bit stream. Build an inheritance-based DVP list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₀Using A₀Control point affine deformation vector of₀And calculating a first affine CPDVP inheritance candidate by using a six-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₀Using B₀Control point affine deformation vector, B₀And calculating a second affine CPDVP inheritance candidate by using a six-parameter affine model. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). Deforming affine of neighboring blocks of a current blockThe vectors are divided into a first group, a second group and a third group, the CPDVP candidate of the zeroth control point CP0 of the current block is derived from the first group, the CPDVP candidate of the first control point CP1 of the current block is derived from the second group, and the CPDVP candidate of the second control point CP2 of the current block is derived from the third group. Wherein the first group comprises block vectors of neighboring blocks A

Block vector of neighboring block B

Block vector of neighboring block C

The second group comprises block vectors of neighboring blocks D

And block vectors of neighboring blocks E

The third group comprises block vectors of neighboring blocks F

And block vectors of neighboring blocks G

Checking in the search order A through B through C if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 0. Checking in the search order D to E if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 1. Checking in the search order from F to G if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 2.

The set constitutes a first CPDVP construction candidate. And predicting a CPDVP index value CPDVP idx according to the affine deformation vectors of the three control points of the current block, and finding the affine deformation vector prediction CPDVP of the three control points of the current block corresponding to the CPDVP idx in the inherited DVP-based list and the constructed DVP-based list. And adding the affine deformation vector prediction CPDVP of the three control points and the affine deformation vector difference CPDVD of the three control points to obtain affine deformation vectors CPDV of the three control points, using a six-parameter affine model, the affine deformation vectors CPDV of the three control points of the current block and the spatial positions of the three control points to deduce the affine deformation vectors of the two sub-blocks in the current block, and performing prediction compensation in the current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the process is integer pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the sub-block used in all the areas is integer pixel precision when the current image is decoded. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 15

obtaining affine deformation vector prediction CPDVP index values CPDVP idx of two control points of the current block, affine deformation vectors CPDV of the two control points and affine deformation vector difference values CPDVD between the two affine deformation vector prediction CPDVPs from the bit stream. Build an inheritance-based DVP list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₁Using A₁Control point affine deformation vector of₁The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₁Using B₁Control point affine deformation vector, B₁And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). Neighbor of current blockThe affine transformation vectors of the near block are divided into a first group, a second group and a third group, the CPDVP candidate of the zeroth control point CP0 of the current block is derived from the first group, the CPDVP candidate of the first control point CP1 of the current block is derived from the second group, and the CPDVP candidate of the second control point CP2 of the current block is derived from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

the set constitutes a third CPDVP construction candidate. Predicting CPDVP index value CPDVP idx according to affine deformation vectors of two control points of current block in inheritance-based DVP list sumAnd predicting the CPDVP based on the affine deformation vectors of the two control points of the current block corresponding to the CPDVP idx found in the constructed DVP list. Adding the affine deformation vector prediction CPDVP of the two control points and the affine deformation vector difference value CPDVD of the two control points to obtain affine deformation vectors CPDV of the two control points, using a four-parameter affine model, the affine deformation vectors CPDV of the two control points of the current block and the space positions of the two control points to deduce the affine deformation vectors of the two sub-blocks in the current block, and performing prediction compensation in the current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. In performing prediction compensation, samples in the current picture include samples required for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded sample is used as the not yet decoded sample. The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the subblock involved in the process is 1/16 sub-pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the subblock used in all regions when decoding the current image is 1/16 sub-pixel precision. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 16

and obtaining affine deformation vector prediction CPDVP index values CPDVP idx of two control points of the current block and affine deformation vector differences CPDVD between the affine deformation vectors CPDV of the two control points and the three affine deformation vector prediction CPDVPs from the bit stream. Build an inheritance-based DVP list: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₁Using A₁Control point affine deformation vector of₁The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₂Using B₂Control point affine deformation vector, B₂And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. Constructing a DVP list based on the construction: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

the set constitutes a third CPDVP construction candidate. And predicting a CPDVP index value CPDVP idx according to the affine deformation vectors of the two control points of the current block, and finding the affine deformation vector prediction CPDVP of the two control points of the current block corresponding to the CPDVP idx in the inherited DVP-based list and the constructed DVP-based list. Adding the affine deformation vector prediction CPDVP of the two control points and the affine deformation vector difference value CPDVD of the two control points to obtain affine deformation vectors CPDV of the two control points, using a four-parameter affine model, the affine deformation vectors CPDV of the two control points of the current block and the space positions of the two control points to deduce the affine deformation vectors of the two sub-blocks in the current block, and performing prediction compensation in the current image according to the affine deformation vectors of the two sub-blocks to obtain prediction samples of the two sub-blocks. Samples in the current picture include samples needed for prediction but not yet decoded, and the top nearest decoded sample of the not yet decoded samples is used as the not yet decoded sampleThe samples are decoded. The spatial accuracy of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV, and the affine deformation vector of the subblock involved in this process is 1/16 sub-pixel accuracy, and the spatial accuracy of the affine deformation vector CPDV of the control point and the affine deformation vector of the subblock used in all regions when decoding the current image is different, including integer pixel accuracy and 1/16 sub-pixel accuracy. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like. The flow chart is shown in figure 4.

Example 17

The present example provides an intra prediction apparatus for an image, specifically including:

the device diagram is shown in figure 7.

Affine deformation information analysis module 101: the affine deformation vector CPDV of the two control points of the current block is parsed and output from the input bitstream,

the control point affine deformation vector calculation block 102: the affine deformation vectors CPDV of the two control points of the current block output by the affine deformation information analysis module 101 are input, the affine deformation vectors CPDV of the two control points of the current block are subjected to identity processing, the affine deformation vectors CPDV of the two control points of the current block are output,

sub-block affine deformation vector calculation block 103: the spatial positions of the two control points of the current block and the affine deformation vectors CPDV of the two control points of the current block output by the control point affine deformation vector calculation module 102 are input, the four-parameter affine model is used for calculating and outputting the affine deformation vectors which are different and are of the two sub blocks in the current block,

the predicted sample acquisition module 104: the input is the affine deformation vectors of the two sub-blocks within the current block and the samples in the current image, which are output by the sub-block affine deformation vector calculation block 103, the prediction samples of the two sub-blocks are derived and output,

in the predicted sample acquisition module 104, the undecoded sample generation module 106 is included: the samples in the current picture include samples required for prediction but not yet decoded, the input is all decoded samples in the same row as the not yet decoded samples in the current picture, the average value of all decoded samples is calculated, the average value is taken as the not yet decoded samples and output,

in the prediction sample obtaining module 104, the prediction sample deriving module 107 is further included: the input is the affine deformation vector of the sub-block, the decoded samples in the current image, and the un-decoded samples output by the un-decoded sample generation module 106, and the prediction samples of the sub-block are derived and output.

The spatial precision of the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the device is integer pixel precision, and the spatial precision of the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to all the areas is integer pixel precision when the current image is decoded. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like.

Example 18

the device diagram is shown in figure 7.

in the predicted sample acquisition module 104, the undecoded sample generation module 106 is included: the samples in the current picture include samples required for prediction but not yet decoded, the input is all decoded samples in the same column of the current picture as the not yet decoded samples, the average value of all decoded samples is calculated, the average value is taken as the not yet decoded samples and output,

Example 19

the device diagram is shown in figure 7.

in the predicted sample acquisition module 104, the undecoded sample generation module 106 is included: the samples in the current image comprise samples which are needed by prediction but not decoded yet, the input is all decoded samples in the current image which are in the same row and the same column with the samples not decoded yet, the generated samples are calculated by using all decoded samples according to the 45-degree prediction direction by using an intra-direction prediction method, the generated samples are taken as the samples not decoded yet and output,

Example 20

the device diagram is shown in figure 7.

in the predicted sample acquisition module 104, the undecoded sample generation module 106 is included: the samples in the current picture include samples required for prediction but not yet decoded, the input is the current picture, the leftmost sample position of the picture in the same row as the sample not yet decoded is used as a search starting point, searching is carried out in the right (upper, lower) direction, the first decoded sample is selected as the generated sample, if the search range in the right (upper, lower) direction exceeds the picture range and the decoded sample is not searched yet, searching is carried out in the lower (left, right) direction by using the uppermost sample position of the picture in the same row as the sample not yet decoded as a search starting point, the first decoded sample is selected as the generated sample and output,

Example 21

the device diagram is shown in figure 7.

in the predicted sample acquisition module 104, the undecoded sample generation module 106 is included: the samples in the current picture include samples required for prediction but not yet decoded, the input is the current picture, the nearest sample position to the left of the not yet decoded samples is a search starting point, a search is performed in the left direction, the first decoded sample is selected as the generated sample, if the search range in the left direction exceeds the picture range and the decoded sample is not yet searched, the nearest sample position above (to the right, below) the not yet decoded sample is used as the search starting point, a search is performed in the upward (to the right, below) direction, and the first decoded sample is selected as the generated sample. And taking the generated sample as the undecoded sample and outputting the undecoded sample.

Example 22

the device diagram is shown in figure 7.

in the predicted sample acquisition module 104, the undecoded sample generation module 106 is included: the samples in the current image comprise samples which are needed by prediction but not decoded yet, the input is the current image, the top nearest sample position of the sample which is not decoded is used as a search starting point, the search is carried out in the top direction, the first decoded sample is selected as the generated sample, if the search range in the top direction exceeds the image range, the decoded sample is not searched yet, the left (right, lower) nearest sample position of the sample which is not decoded yet is used as the search starting point, the search is carried out in the left (right, lower) direction, and the first decoded sample is selected as the generated sample. And taking the generated sample as the undecoded sample and outputting the undecoded sample.

Example 23

the device diagram is shown in figure 7.

Affine deformation information analysis module 101: the affine deformation vector CPDV of the three control points of the current block is parsed and output from the input bitstream,

the control point affine deformation vector calculation block 102: the affine deformation vectors CPDV of the three control points of the current block output by the affine deformation information analysis module 101 are input, the affine deformation vectors CPDV of the three control points of the current block are subjected to identity processing, and the affine deformation vectors CPDV of the three control points of the current block are output

Sub-block affine deformation vector calculation block 103: the spatial positions of the three control points of the current block and the affine deformation vectors CPDV of the three control points of the current block output by the control point affine deformation vector calculation module 102 are input, the six-parameter affine model is used for calculating and outputting the affine deformation vectors which are different and are of the two sub-blocks in the current block,

in the predicted sample acquisition module 104, the undecoded sample generation module 106 is included: the samples in the current picture comprise samples that are needed for prediction but not yet decoded, the input being the decoded sample that is closest to the left of the not yet decoded sample in the current picture, the decoded sample being taken as the generated sample. And taking the generated sample as the undecoded sample and outputting the undecoded sample.

Example 24

the device diagram is shown in figure 7.

Affine deformation information analysis module 101: resolving and outputting affine deformation vector predictions CPDVPs of three control points of the current block and affine deformation vector differences CPDVDs between the affine deformation vectors CPDV of the three control points and the three affine deformation vector predictions CPDVPs from an input bit stream,

the control point affine deformation vector calculation block 102: affine deformation vector predictions CPDVPs of the three control points of the current block, output by the affine deformation information analyzing module 101, affine deformation vectors CPDVPs of the three control points and affine deformation vector difference values CPDVD between the three affine deformation vector predictions CPDVPs are input, the affine deformation vector predictions CPDVPs of the three control points and the affine deformation vector difference values CPDVD of the three control points are added to obtain affine deformation vectors CPDV of the three control points, the affine deformation vectors CPDV of the three control points of the current block are output,

Example 25

the device diagram is shown in figure 8.

Affine deformation information analysis module 101: resolving and outputting affine deformation vector prediction CPDVP index values CPDVP idx of two control points of the current block from an input bit stream,

the control point affine deformation vector calculation block 102: inputting affine deformation vector prediction CPDVP index values CPDVP idx of two control points of a current block output by an affine deformation information analysis module 101, constructing an inheritance-based DVP list and a construction-based DVP list, finding affine deformation vector prediction CPDVP of the two control points of the current block corresponding to the CPDVP idx in the inheritance-based DVP list and the construction-based DVP list, carrying out identity processing on the affine deformation vector prediction CPDVP of the two control points, taking the affine deformation vector prediction DPDPDPDPDPDs of the two control points as the affine deformation vectors CPDV of the two control points, and outputting the affine deformation vectors CPDV of the two control points of the current block.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 108: the input is decoded information, and an inheritance-based DVP list is constructed: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₁To A₀Searching ofThe index sequentially checks the neighboring blocks in the left block group if the first block that satisfies the condition is A₁Using A₁Control point affine deformation vector of₁The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₁To B₀To B₂If the first block satisfying the condition is B₁Using B₁Control point affine deformation vector, B₁And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. An inheritance-based DVP list, comprised of a first affine CPDVP inheritance candidate and a second affine CPDVP inheritance candidate, is output.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 109: the input is decoded information, and a DVP list based on the construction is constructed: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group only comprising neighborsBlock vector of block D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

the set constitutes a third CPDVP construction candidate. Outputting a construction-based DVP list consisting of the first CPDVP construction candidate, the second CPDVP construction candidate and the third CPDVP construction candidate.

Example 26

the device diagram is shown in figure 8.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 108: the input is decoded information, and an inheritance-based DVP list is constructed: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₀Using A₀Control point affine deformation vector of₀The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₀Using B₀Control point affine deformation vector, B₀And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. An inheritance-based DVP list, comprised of a first affine CPDVP inheritance candidate and a second affine CPDVP inheritance candidate, is output.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 109: the input is decoded information, and a DVP list based on the construction is constructed: as shown in FIG. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, let us say that the neighboring block A is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1),the neighboring block F is a block including samples of coordinates (-1, H-1), and the neighboring block G is a block including samples of coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

Example 27

the device diagram is shown in figure 8.

Affine deformation information analysis module 101: resolving and outputting affine deformation vector prediction CPDVP index values CPDVP idx of three control points of the current block from an input bit stream,

the control point affine deformation vector calculation block 102: inputting affine deformation vector prediction CPDVP index values CPDVP idx of three control points of a current block output by an affine deformation information analysis module 101, constructing an inheritance-based DVP list and a construction-based DVP list, finding affine deformation vector prediction CPDVP of the three control points of the current block corresponding to the CPDVP idx in the inheritance-based DVP list and the construction-based DVP list, carrying out identity processing on the affine deformation vector prediction CPDVP of the three control points, taking the affine deformation vector prediction DPDPDPDPDPDs of the three control points as the affine deformation vectors CPDV of the three control points, and outputting the affine deformation vectors CPDV of the three control points of the current block.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 108: the input is decoded information, and an inheritance-based DVP list is constructed: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (2) checks neighboring blocks in the left block group, and if there are no blocks that satisfy the condition, takes the CPDVP formed by zero values as the first affine CPDVP inheritance candidate. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If there are no blocks that satisfy the condition, the CPDVP formed by the zero values is taken as a second affine CPDVP inheritance candidate. An inheritance-based DVP list, comprised of a first affine CPDVP inheritance candidate and a second affine CPDVP inheritance candidate, is output.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 109: the input is decoded information, and a DVP list based on the construction is constructed: as shown in FIG. 6, the size of the current block is WxH and whenIn the case where the x-component and y-component of the top left sample position of the preceding block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate. A construction-based DVP list consisting of the first CPDVP construction candidates is output.

Example 28

the device diagram is shown in figure 8.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 109: the input is decoded information, and a DVP list based on the construction is constructed: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises block vectors of neighboring blocks A

Block vector of neighboring block B

Block vector of neighboring block C

The second group comprises block vectors of neighboring blocks D

And block vectors of neighboring blocks E

The third group comprises block vectors of neighboring blocks F

And block vectors of neighboring blocks G

Checking in the search order A through C through B if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 0. Checking in the search order E to D if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 1. Checking in the search order G through F if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 2.

The set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

Example 29

the device diagram is shown in figure 8.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 109: the input is decoded information, and a DVP list based on the construction is constructed: as shown in FIG. 6, in the large of the current blockLet us say that neighboring block a is a block of samples including coordinates (-1, -1), neighboring block B is a block of samples including coordinates (0, -1), neighboring block C is a block of samples including coordinates (-1,0), neighboring block D is a block of samples including coordinates (W-1, -1), neighboring block E is a block of samples including coordinates (W, -1), neighboring block F is a block of samples including coordinates (-1, H-1), and neighboring block G is a block of samples including coordinates (-1, H) in the case where WxH is small and the x-component and y-component of the upper-left sample position of the current block are 0. The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises block vectors of neighboring blocks A

Block vector of neighboring block B

Block vector of neighboring block C

The second group comprises block vectors of neighboring blocks D

And block vectors of neighboring blocks E

The third group comprises block vectors of neighboring blocks F

And block vectors of neighboring blocks G

Checking in the search order A through B through C if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 0. Checking in the search order D to E if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 1. Checking in the search order from F to G if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 2.

The set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

Example 30

the device diagram is shown in figure 8.

Affine deformation information analysis module 101: resolving and outputting affine deformation vector prediction CPDVP index values CPDVP idx of the three control points of the current block and affine deformation vector difference values CPDVD between the affine deformation vectors CPDV of the three control points and the three affine deformation vector predictions CPDVP from an input bit stream,

the control point affine deformation vector calculation block 102: the affine deformation vector prediction CPDVP index value CPDVP idx of the current three control points output by the affine deformation information analysis module 101, the affine deformation vector CPDV of the three control points and the affine deformation vector difference CPDVD between the three affine deformation vector predictions CPDVP are input, an inheritance-based DVP list and a construction-based DVP list are built, the affine deformation vector prediction CPDVP of the current three control points corresponding to the CPDVP idx is found in the inheritance-based DVP list and the construction-based DVP list, the affine deformation vector prediction CPDVP of the three control points and the affine deformation vector difference CPDVD of the three control points are added to obtain the affine deformation vectors CPDV of the three control points, and the affine deformation vectors CPDV of the current three control points are output.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 108: the input is decoded information, and an inheritance-based DVP list is constructed: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₀Using A₀Control point affine deformation vector of₀And calculating a first affine CPDVP inheritance candidate by using a six-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₀Using B₀Control point affine deformation vector, B₀And calculating a second affine CPDVP inheritance candidate by using a six-parameter affine model. An inheritance-based DVP list, comprised of a first affine CPDVP inheritance candidate and a second affine CPDVP inheritance candidate, is output.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 109: the input is decoded information, construct D based on structureVP list: as shown in fig. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, note that the neighboring block a is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), the neighboring block E is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises block vectors of neighboring blocks A

Block vector of neighboring block B

Block vector of neighboring block C

The second group comprises block vectors of neighboring blocks D

And block vectors of neighboring blocks E

The third group comprises block vectors of neighboring blocks F

And block vectors of neighboring blocks G

Checking in the search order A through B through C if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 0. Checking in the search order D to E if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 1. Checking in the search order from F to G if

Is the first block vector to satisfy the condition, will

As CPDVP candidates for CP 2.

Example 31

the device diagram is shown in figure 8.

Affine deformation information analysis module 101: resolving and outputting affine deformation vector prediction CPDVP index values CPDVP idx of two control points of the current block and affine deformation vector differences CPDVDs between the affine deformation vectors CPDV of the two control points and the two affine deformation vector predictions CPDVP from an input bit stream,

the control point affine deformation vector calculation block 102: the method comprises the steps of inputting affine deformation vector prediction CPDVP index values CPDVP idx of two control points of a current block output by an affine deformation information analysis module 101, affine deformation vectors CPDV of the two control points and affine deformation vector difference CPDVD between the two affine deformation vector predictions CPDVP, constructing an inheritance-based DVP list and a construction-based DVP list, finding affine deformation vector prediction CPDVP of the two control points of the current block corresponding to the CPDVP idx in the inheritance-based DVP list and the construction-based DVP list, adding the affine deformation vector prediction CPDVP of the two control points and the affine deformation vector difference CPDVD of the two control points to obtain affine deformation vectors CPDV of the two control points, and outputting the affine deformation vectors CPDV of the two control points of the current block.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 108: the input is decoded information, and an inheritance-based DVP list is constructed: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₁Using A₁Control point affine deformation vector of₁The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂If the first block satisfying the condition is B₁Using B₁Control point affine deformation vector, B₁And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. An inheritance-based DVP list, comprised of a first affine CPDVP inheritance candidate and a second affine CPDVP inheritance candidate, is output.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 109: the input is decoded information, and a DVP list based on the construction is constructed: as shown in FIG. 6, in the case where the size of the current block is WxH and the x-component and y-component of the upper-left sample position of the current block are 0, let us say that the neighboring block A is a block of samples including coordinates (-1, -1), the neighboring block B is a block of samples including coordinates (0, -1), the neighboring block C is a block of samples including coordinates (-1,0), the neighboring block D is a block of samples including coordinates (W-1, -1), and the neighboring block E is a block of samples including coordinates (W-1, -1)Is a block of samples including coordinates (W, -1), the neighboring block F is a block of samples including coordinates (-1, H-1), and the neighboring block G is a block of samples including coordinates (-1, H). The method includes dividing affine deformation vectors of neighboring blocks of a current block into a first group, a second group, and a third group, deriving a CPDVP candidate of a zeroth control point CP0 of the current block from the first group, deriving a CPDVP candidate of a first control point CP1 of the current block from the second group, and deriving a CPDVP candidate of a second control point CP2 of the current block from the third group. Wherein the first group comprises only block vectors of neighboring blocks A

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

the set constitutes a third CPDVP construction candidate. The output is composed of a first CPDVP construction candidate, a second CPDVP construction candidate and a third CPDVP construction candidateBased on the constructed DVP list.

Sub-block affine deformation vector calculation block 103: the spatial positions of the two control points of the current block and the affine deformation vectors CPDV of the two control points of the current block output by the control point affine deformation vector calculation module 102 are input, and the four-parameter affine model is used for calculating and outputting the affine deformation vectors of the two sub-blocks in the current block, which are different from each other.

The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the device is 1/16 sub-pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the sub-block used in all regions when the current image is decoded is 1/16 sub-pixel precision. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like.

Example 32

the device diagram is shown in figure 8.

The control point affine deformation vector calculation module 102 includes an inheritance-based DVP list construction module 108: the input is decoded information, and an inheritance-based DVP list is constructed: as shown in FIG. 5, let A be the lower left adjacent block of the current block₁The left adjacent block of the current block is A₀The adjacent block at the upper left corner of the current block is B₂The neighboring block on the current block is B₀The upper right adjacent block of the current block is B₁. In the lower left corner adjacent block A₁And a left adjacent block A₀In the left block group formed, according to A₀To A₁The search order of (a) checks neighboring blocks in the left block group if the first block that satisfies the condition is a₁Using A₁Control point affine deformation vector of₁The first affine CPDVP inheritance candidate is calculated by using a four-parameter affine model. The adjacent block at the upper left corner of the current block is B₂The adjacent block on the current block is B₀And the upper right adjacent block of the current block is B₁In the upper block group formed, according to B₀To B₁To B₂Search order ofChecking neighboring blocks in the upper block group if the first block satisfying the condition is B₂Using B₂Control point affine deformation vector, B₂And calculating a second affine CPDVP inheritance candidate by using a four-parameter affine model. An inheritance-based DVP list, comprised of a first affine CPDVP inheritance candidate and a second affine CPDVP inheritance candidate, is output.

The second group comprises only block vectors of neighboring blocks D

The third group comprises only block vectors of neighboring blocks F

Will be provided with

As CPDVP candidates for CP0, will

As CPDVP candidates for CP1, will

As a CPDVP candidate for the CP2,

the set constitutes a first CPDVP construction candidate,

the set constitutes a second CPDVP construction candidate,

The spatial precision of the control point affine deformation vector prediction CPDVP, the control point affine deformation vector CPDV and the affine deformation vector of the sub-block related to the device is 1/16 sub-pixel precision, and the spatial precision of the affine deformation vector CPDV of the control point and the affine deformation vector of the sub-block used in all regions when the current image is decoded is different, wherein the spatial precision comprises integer pixel precision and 1/16 sub-pixel precision. The space size of the control point and the sub-block can be a pixel point with the size of 1x1, and can also be a regular or irregular pixel area such as 4x4, 8x8 and the like.

The foregoing is a more detailed description of the present invention that is presented in conjunction with specific embodiments, and the practice of the invention is not to be considered limited to those descriptions. For those skilled in the art to which the invention pertains, several equivalent substitutions or obvious modifications can be made without departing from the spirit of the invention, and the same properties or uses are to be considered as falling within the scope of the invention.

Claims

1. A method for intra prediction of an image, characterized by:

2. The method of claim 1, wherein:

the not yet decoded samples are generated from at least two decoded samples in the current picture adjacent to the not yet decoded sample position.

3. The method according to claim 2, wherein the generation rule of the not-yet-decoded samples is:

and calculating a generated sample by using all decoded samples according to a certain prediction direction by using an intra-frame direction prediction method, and taking the generated sample as the undecoded sample.

4. The method according to claim 1, wherein the generation rule of the not-yet-decoded samples is:

according to a certain search rule, selecting a decoded sample adjacent to the position of the sample which is not decoded in the current image as a generated sample, and using the generated sample as the sample which is not decoded.

5. The method of claim 4, wherein:

the certain search rule comprises that the position, closest to the sample not yet decoded, of a certain direction of the sample not yet decoded is used as a search starting point, the direction is searched, and the first decoded sample is selected as the generated sample; if the previous direction searching range exceeds the image range and the decoded sample is not searched yet, the position of the other direction of the not-yet-decoded sample, which is nearest to the not-yet-decoded sample, is taken as a searching starting point, the other direction is searched, and the first decoded sample is selected as the generated sample.

6. An intra prediction apparatus for an image, characterized in that:

affine deformation information analysis module: analyzing and outputting affine deformation information of the current block from an input bit stream;

a control point affine deformation vector calculation module: calculating and outputting affine deformation vectors of at least two control points of the current block by using the input affine deformation information of the current block;

a sub-block affine deformation vector calculation module: calculating and outputting affine deformation vectors which are different and are of at least two sub-blocks in the current block from the input affine deformation vector of the control point of the current block and the position of the control point of the current block by using an affine model;

a predicted sample acquisition module: deriving and outputting a prediction sample of at least one sub-block from the input affine deformation vector of the sub-block and a sample in a current image;

the prediction sample acquisition module comprises an undecoded sample generation module and a prediction sample derivation module;

the undecoded sample generation module: if the samples in the current image comprise samples which are needed by prediction but not decoded yet, inputting decoded samples which comprise positions adjacent to the samples which are not decoded yet in the current image, and generating and outputting the samples which are not decoded yet according to a generating rule;

7. The apparatus according to claim 6, wherein:

in the un-decoded sample generation module, the input is at least two decoded samples adjacent to the un-decoded sample position in the current picture.

8. The apparatus according to claim 7, wherein the generation rule of the not-yet-decoded samples is:

9. The apparatus according to claim 6, wherein the generation rule of the not-yet-decoded samples is:

10. The apparatus for intra prediction of an image according to claim 9, wherein: