KR20210124824A - A method and an apparatus for processing a video signal - Google Patents

Abstract

The present invention provides a method and device for predictive encoding/decoding using a block vector. The purpose of the present invention is to improve the coding efficiency of a video signal.

Description

A METHOD AND AN APPARATUS FOR PROCESSING A VIDEO SIGNAL

The present invention relates to a video signal processing method and apparatus.

The video image is compression-encoded by removing spatial and temporal redundancy and inter-view redundancy, which may be transmitted through a communication line or stored in a form suitable for a storage medium.

An object of the present invention is to improve the coding efficiency of a video signal.

In order to solve the above problems, the present invention provides a method and apparatus for predictive encoding/decoding using a block vector.

The video signal processing method and apparatus according to the present invention can improve video coding efficiency by supporting prediction using block vectors.

Recently, ultra-high-resolution video is the core of not only digital broadcasting but also streaming services such as Netflix and YouTube. In addition to the existing 2D images, VR and 3D image services are being commercialized, and the above image services can be used not only on digital TVs but also on mobile devices such as smartphones. What these video services have in common is that the service is impossible without the application of video compression. In the case of 1080p@60Hz, which can be called Full-HD, a 1920x1080 screen must be transmitted 60 times per second. Like 3D video, double data is required to deliver information to both eyes, and ultra-high-resolution video services such as 4K (4096x2048) and 8K (8192x4096) require one screen to be transmitted more than 120 times per second. Huge amount of data is generated compared to HD. In order to handle such data, technologies in various fields such as communication bandwidth and image compression technology are required.

The inter prediction technology is a technology that searches for a prediction block most similar to the current block in a reference image by using a vector, and intra block copy (IBC) searches for a prediction block most similar to the current block by using a vector in the current image. is a technique to A vector used in the inter prediction technique is referred to as a motion vector (MV), and a vector used in IBC is referred to as a block vector (BV). In the above two methods, merging can be performed using the restored vectors around the current block, respectively, or it can be used as a predicted value of the vector used in the current block.

One. vector derivation method

Duplicate data existing in the screen may be removed through IBC. A prediction block generated by IBC is generated by BV. In the current image, after performing a search using BV in a previously reconstructed region (reference region) before the block to be encoded (current block), a block most similar to the current block is generated and a difference value with the current block is generated.

The following figure 1 is an example showing the search process using BV in the current image.

Figure 1

Alternatively, unlike Figure 1, it is also possible to store the reference region in a temporary buffer of a preset size and search for a BV within the region. In this case, the temporary buffer may include restored consecutive blocks.

Information indicating the size of the buffer may be signaled through a bitstream. Alternatively, the size of the buffer may be determined based on the size of the coding tree unit. The buffer may be initialized for each predetermined unit. The predetermined unit may be at least one of a slice, a tile, a row of N coding tree units, or N coding tree units. Here, N may be 1 or a natural number of 1 or more.

BV(x, y) optimally selected from among several temporary BVs (x', y') generated during search is encoded and transmitted to a decoder. In this case, the BV may be encoded using decoded BVs existing around the current block. For example, an index specifying a neighboring block having the same BV as the current block may be encoded, or the BV of the neighboring block may be set as a BV prediction value, and a difference value between the BV and the BV prediction value of the current block may be transmitted. The following Figure 2 is a diagram for explaining a reference position of a decoded BV that can be used to encode a BV of a current block.

Figure 2

In Figure 2, LB means the position of the lower-left pixel in the block to be currently encoded, and RT means the position of the upper-right pixel in the block. For example, A1 denotes a position of a restoration pixel that exists immediately to the left of LB, and B1 denotes a position of a restoration pixel that exists immediately above RT.

After selecting a BV existing in a block including neighboring reconstructed pixels as a predicted value according to a preset priority, a difference value with the BV derived from the block to be currently encoded may be encoded and transmitted to the decoder. Alternatively, the BV present in the block including the neighboring reconstructed pixels may be set as the BV of the block to be currently encoded.

In this case, a BV candidate list may be constructed, and an index may indicate which BV is used to derive the BV of the block to be currently encoded. For example, if it is assumed that the size of the BV candidate list is 2, a BV candidate may be derived based on a block including a sample at position B1 and a block including a sample at position A1. BV candidates may be added to the BV candidate list in the order of B1->A1 or A1->B1.

Alternatively, depending on the size of the BV candidate list, BV candidates may be derived from blocks at positions A0, B0, or B5, or BV candidates may be derived from blocks located at other positions.

The number of available candidate blocks may be set differently according to the size of the current block. For example, when the size of the current block is 16 or less (eg, 4x4, 2x8, 8x2, etc.), neighboring blocks may be set not to be used as candidate blocks. On the other hand, when the size of the current block is greater than 16, blocks at positions B1 and A1 may be set to be used as candidate blocks.

Alternatively, a specific region may be designated when constructing the BV candidate list, and the candidate block may be inserted into the candidate list only when it exists outside the specific region. If the candidate block exists inside a specific region, it is determined that the candidate block is unavailable and may not be inserted into the candidate list.

Alternatively, in contrast to the above, if the candidate block exists outside the specific region, it may be determined as unavailable and not inserted into the candidate list. If the candidate block exists inside a specific area, it may be inserted into the candidate list.

That is, the availability of the neighboring block may be determined based on whether the current block and the neighboring block belong to the same specific region.

Here, the specific region may be a parallel processing region (or merge processing region), a CTU, a CTU row, or a slice.

In the following example, for convenience of explanation, it is assumed and described that it is possible to insert into the candidate list only when a surrounding location referring to the BV exists outside a specific region.

Figure 3 and Equation (1) below show examples related to this.

Figure 3

Nflag = { ( xPb>> Log2BlkLevel ) == ( xNb>> Log2BlkLevel ) } &&

{ ( yPb >> Log2BlkLevel ) == ( yNb >> Log2BlkLevel ) } (1)

In Figure 3, (xPb,yPb) means the position of the upper leftmost pixel in the current block. (xNb, yNb) means the pixel position of the neighboring block, and according to Figure 2, (xNb, yNb) may be A0 to A4 or B0 to B4. Therefore, in the example of Figure 3, (xNb, yNb) can be A1 or B1.

In Equation (1), Log2BlkLevel is a variable indicating the size of a specific area. Log2BlkLevel can be derived by taking _{Log 2} to the size of a specific region. Information indicating the size of a specific region may be signaled through a bitstream. For example, it may be encoded through the upper header and transmitted to the decoder. Alternatively, the size of a specific region may be predefined in the encoder and the decoder.

Figure 4 below shows an example of Figure 3 and Equation (1).

Figure 4

In the example below, Nflag (N is A or B) indicates the availability of blocks around N positions.

For example, when the value of Log2BlkLevel is 2 and (xPb,yPb) is (12,12), Aflag is set to true by Equation (1) and BV existing at position A can be inserted into the candidate list. . However, if the value of Log2BlkLevel is set to 3, Aflag is set to false and the BV existing at position A is not inserted into the candidate list, and the corresponding position is judged as unreferenceable. Similarly, the same restrictions as the A location may be applied to the B location.

When the number of BV candidates included in the BV candidate list is equal to or less than the threshold, the BV candidates included in the temporary vector candidate list may be included in the BV candidate list. The temporary vector candidate list may include BV candidates derived based on blocks encoded/decoded with IBC before the current block. When a BV candidate included in the temporary vector candidate list is the same as a BV candidate previously inserted into the BV candidate list, the corresponding BV candidate may be set not to be added to the BV candidate list. One of the BV candidates included in the candidate list may be used as a prediction value for the BV of the current block. In addition, the residual BV of the current block may be encoded by differentiating the BV of the current block and the BV predicted value. In this case, an index indicating which BV among BVs stored in the list is used as a prediction value may also be encoded and transmitted to the decoder. This method is called an adaptive block vector prediction (ABVP) method.

Alternatively, the BV stored in the list may be set as the BV of the current block. In this case, an index indicating which BV among the BVs stored in the list is used as the BV of the current block may be encoded and transmitted to the decoder. This method is called the BV merging method.

Alternatively, when using the ABVP method or the merging method, only one candidate may be used and the index may not be encoded. For example, only BV candidates existing in the first of the BV candidate list may be used, or candidate blocks may be searched in a predetermined order, and a BV candidate of the first found available candidate block may be used. In this case, the index may not be encoded.

Alternatively, Equation (1) above may be changed to Equation (2) below.

Nflag = { (xPb>> Log2BlkLevel ) == ( xNb>> Log2BlkLevel ) } ||

{ ( yPb >> Log2BlkLevel ) == (yNb >> Log2BlkLevel ) } (2)

Alternatively, Equation (1) or Equation (2) may be used according to the size or shape of the block. For example, if the current block is a square, Equation (1) may be used, and if the current block is a rectangle, Equation (2) may be used.

Alternatively, when the current block is a rectangle, a similar constraint may be applied in the direction of the larger side of the two sides. For example, when a block has a horizontal length of 16 and a vertical length of 8, the following Equation (3) may be applied.

Nflag = (xPb>> Log2BlkLevel ) == ( xNb>> Log2BlkLevel ) (3)

Conversely, when the horizontal length of the block is 8 and the vertical length is 16, the following Equation (4) can be applied.

Nflag = (yPb>> Log2BlkLevel ) == ( yNb>> Log2BlkLevel ) (4)

Alternatively, when the current block is a rectangle, it is also possible to apply a similar constraint in the direction of the smaller of the two sides, as opposed to the above.

Alternatively, when the current block is a rectangle, Equation (2) may be applied according to the ratio of the lengths of the two sides. For example, the constraint like Equation (2) is applied only when the ratio of the lengths of the two sides is 4:1 or more, otherwise it may not be applied. This ratio may have a fixed value in the encoder and the decoder, or it is also possible to encode the ratio through a higher header and transmit it to the decoder.

The conditions for the above block length or ratio can be specified differently for each area (slice, tile, subpicture, etc.) in which parallel processing can be performed.

Alternatively, the value of Log2BlkLevel may be transmitted in SPS or PPS, or may be transmitted by encoding a different value for each region (slice, tile, subpicture, etc.) in which parallel processing can be performed.

Alternatively, it is also possible to separately encode the Log2BlkLevel value corresponding to each of the width and length of the block. For example, the value of Log2BlkLevel applied to the horizontal length may be encoded as 2, and the value of Log2BlkLevel applied to the vertical length may be 3.

In the above example, the case of IBC is described as an example for convenience of explanation, but it is equally applicable to the MV merging and AMVP methods used in inter prediction. That is, when inter prediction is applied, availability as an MV candidate may be determined based on whether the neighboring block belongs to the same specific region as the current block.

Several blocks may exist in a specific region caused by the above restrictions, and a temporary vector list can be constructed using vectors of these blocks. Hereinafter, the temporary vector list will be described, and the relationship with the specific region will be described.

The temporary vector list may be constructed in various ways. For example, from the block present at the first position of the image, slice, coding tree unit line, or coding tree unit including the block to be currently encoded, all vectors used up to immediately before the block to be encoded are stored in the temporary vector list. do.

Alternatively, after setting the number N of vectors to be stored in the list, the N vectors are stored in the list.

Indices can be assigned to vectors according to the order in which they are inserted into the temporary vector list. For example, it may have a higher priority as it is closer to the block to be currently encoded, and may have a lower priority as it is further away from the block to be currently encoded. That is, when the encoding order of the first block is earlier than the encoding order of the second block, the vector of the first block may have a higher priority than the vector of the second block. Indices may be assigned to each vector according to priorities, and for example, indices may be assigned in ascending order in the order of priorities or inverse order of priorities.

According to this priority, when storing the vector of the most recently coded block, if N vectors are already stored, the vector of the first coded block is deleted, and the vector of the most recently coded block is deleted. can be saved.

If the same vector as the vector of the most recently coded block already exists in the list, the vector of the most recently coded block may not be added to the list.

Alternatively, when a vector identical to the vector of the most recently coded block already exists in the list, the vector identical to the vector may be deleted and the vector of the most recently coded block may be added to the list.

When the size of the reconstructed block is smaller than the threshold, the vector of the block may not be added to the temporary vector list. For example, when the size of the block is 4x4 or less, the vector of the block may not be added to the temporary vector list.

The temporary vector candidate list may be initialized in a predetermined unit. The predetermined unit may be at least one of a slice, a tile, N CTU rows, N CTUs, or a parallel processing region. N may be 1 or an integer greater than 1.

Alternatively, in the parallel processing structure, a method of separately constructing a temporary vector list for each region to be processed in parallel is also possible.

Alternatively, when a block is included in a parallel processing region, the vector of the block may not be added to the temporary vector list. Alternatively, only a vector of a specific location block among blocks included in the parallel processing region may be added to the temporary vector list. The specific position may be a block including a sample in the upper-left, upper-right, lower-left, lower-right, or center position in the parallel processing region.

Alternatively, after restoration of all blocks in the parallel processing region is completed, vectors of blocks in the parallel processing region may be added to the temporary vector list.

Alternatively, a method of separately configuring a vector list for each CTU row of a region is also possible. In this case, when the vector list is separately provided for each region in which parallel processing is performed, the number of vectors stored in the vector list may be very small in the initial part of the region. Therefore, it is also possible to fill in a preset initial vector instead of filling in a vector from the beginning for each region in which parallel processing is performed. For example, the preset initial vector may be a vector derived from the entire image, not a vector derived for each block. In this case, each value of the vector derived from the entire image may be encoded through a higher header along with the initial number of vectors.

In a specific region generated by Equations (1) to (4), a plurality of blocks may exist according to Log2BlkLevel. In this case, it is possible to store only vectors existing in a block at a predetermined position within a specific area in the temporary vector list. For example, it is possible to store the vector of the first block in the scan order or the vector of the last block in the scan order within a specific area in the temporary vector list.

Alternatively, vectors that can be stored in the temporary vector list within a specific region may include only vectors existing in blocks having a size greater than or equal to a preset size. This size may be encoded and transmitted to a decoder through an upper header. Alternatively, it may be determined according to the shape of the block. For example, a method such that only vectors existing in a square-shaped block are stored in the temporary vector list is possible. Conversely, it is also possible to store only vectors existing in a rectangular block in the temporary vector list.

Alternatively, after a specific area is set by one of Equations (1) to (4), if it is determined that a neighboring candidate exists within a specific area and is not available, the specific area is assumed as one block and the drawing is based on the specific area 2, a method of setting neighboring candidates is also possible. These candidates can be used for a merging method or vector prediction of a current block existing in a specific region. In this case, all blocks within a specific region may have a common neighboring candidate. In this case, the above method may be applied to a size greater than or less than a predetermined specific area.

Claims (1)

Prediction method using block vectors.