KR20230148506A - A method and an apparatus for processing a video signal - Google Patents

Abstract

The present invention provides an encoding/decoding method based on intra prediction of a video signal and an apparatus therefor.

Description

Video signal processing method and apparatus {A METHOD AND AN APPARATUS FOR PROCESSING A VIDEO SIGNAL}

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

Video images are compressed and encoded by removing spatial-temporal redundancy and inter-view redundancy, and can be transmitted through communication lines or stored in a format suitable for storage media.

The present invention seeks to improve the coding efficiency of video signals.

In order to solve the above problems, a method of encoding/decoding a video signal through intra prediction and a device therefor are provided.

The video signal processing method and device according to the present invention can improve video encoding/decoding efficiency.

Recently, ultra-high resolution video is the core of not only digital broadcasting but also streaming services such as Netflix and YouTube. Moreover, in addition to existing 2D video, VR and 3D video services are being commercialized, and the above video 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 they are impossible without applying 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, twice as much data is required to transmit 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, so full- A huge amount of data is generated compared to HD. In order to handle such data, technologies in various fields such as communication bandwidth and video compression technology are required.

One. predictive technology

Forecasting generates a predicted value corresponding to the original value and then creates a residual value by differentiating the two values. Through this process, duplicate data from the original value is removed. In the encoder, the image is divided into blocks and encoding is performed. Accordingly, prediction can be performed on a divided block basis. When a prediction block is generated by the performed prediction, a residual block can be generated by differentiating it from the original block to be encoded. Conversely, in the decoder, the block is restored by adding the prediction block to the restored residual block. This restored block is called a restored block, and if lossless coding has been performed in the encoder, the restored block and the original block have the same value. On the other hand, if lossy coding was performed in the encoder, the restored block may have a different value similar to the original block.

Prediction serves to reduce the redundancy of data to be encoded. The following figure shows an example of intra prediction.

Figure 1

As shown in the picture above, intra-screen prediction has modes 0 to 66. At this time, mode 0 is also called planar mode, mode 1 is DC mode, and modes 2 to 66 are also called directional modes.

Under the planar mode assigned to mode 0, prediction samples can be generated as shown in Figure 2 below.

Figure 2

In the figure above, T and L are examples of surrounding reference samples used when generating prediction blocks in planar mode. T represents the reference sample located in the upper right corner, and L represents the reference sample located in the lower left corner. P1 is a prediction sample for the horizontal direction. P1 can be generated by linear interpolation of T and a reference sample located at the same location on the Y axis as P1. P2 is a prediction sample for the vertical direction. P2 can be generated by linear interpolation of L and a reference sample located at the same position on the X-axis as P2. After that, the final prediction sample is generated by adding the weights of P1 and P2 using the following equation (1).

(One)

weight here and When determining , the horizontal and vertical length of the current block can be considered. Weighted according to the horizontal and vertical length of the current block and may have the same value or may have different values. If the width and height of the block are the same, the weight and By setting , the prediction sample can be set to the average of P1 and P2. If one side of the block is longer than the other, a lower value may be applied to the weight corresponding to the longer side. Alternatively, it is also possible to apply a higher weight to the weight corresponding to the long side.

We explain how to generate a prediction sample under DC mode assigned to mode 1 using the example in Figure 3.

Figure 3

As shown in Figure 3, after calculating the average value of surrounding reference samples, the calculated value is set to all prediction samples in the prediction block. Reference samples include top reference samples and left reference samples. Depending on the type of block, the average value can be calculated using only the top reference samples or the left reference samples. For example, when the horizontal length of a block is greater than the vertical length, or when the ratio between the horizontal and vertical lengths of a block is greater than (or less than) a predefined value, the average value can be calculated using only the upper reference samples. On the other hand, if the horizontal length of the block is smaller than the vertical length, or if the ratio between the horizontal and vertical lengths of the block is less than (or more than) a predefined value, the average value can be calculated using only the left reference samples.

In the directional mode, projection is performed in the direction of the reference sample according to the angle of each directional mode. If a reference sample exists at that location, the reference sample is set as the prediction sample. If there is no reference sample at that location, interpolation is performed using surrounding reference samples, and the interpolated value is set as the prediction sample. Figure 4 below shows an example of this.

Figure 4

In the above example, in the case of prediction sample B, the reference sample exists when projected from the corresponding location in the direction of the reference sample according to the angle. Therefore, the reference sample is set as the prediction sample. In the case of prediction sample A, there is no reference sample at the integer position when projected from that location toward the reference sample according to the angle. Therefore, in this case, interpolation is performed using surrounding integer position reference samples, and then the interpolated value (fractional position reference sample) is set as the prediction sample.

Information about the intra-screen prediction mode applied to the block can be explicitly signaled through the bitstream. At this time, information about the intra-screen prediction mode may be generated based on the intra-screen prediction mode list. Here, a plurality of intra-prediction mode candidates may exist in the intra-prediction mode list. The intra-prediction mode candidate may be derived based on the intra-prediction mode of a neighboring block neighboring the current block.

For example, when signaling the intra-screen prediction mode using the intra-screen prediction mode list, one of the intra-screen prediction mode candidates is selected as the predicted value of the intra-screen prediction mode, and the difference value between the intra-screen prediction mode and the predicted value of the current block It is possible to encode and signal. At this time, an index indicating an intra-prediction mode candidate selected as a prediction value in the intra-prediction mode list can be encoded and signaled to the decoder.

Alternatively, it may be signaled whether an intra-prediction mode candidate identical to the intra-prediction mode of the current block exists. As an example, a 1-bit flag may indicate whether a candidate for the same intra-screen prediction mode as the current blog intra-screen prediction mode exists. If it is determined to exist, which intra-prediction mode candidate in the intra-prediction mode list is the same as the intra-prediction mode of the current block is indicated with an index, and then the index is signaled to the decoder. If it is determined that it does not exist, the indices are reassigned to the remaining intra-prediction modes, excluding the intra-prediction mode candidates existing in the intra-prediction mode list, and then the index corresponding to the intra-prediction mode of the current block is signaled. can do.

2. Line-based coding method

The above-mentioned intra-screen prediction is applied on a block basis. Therefore, as the distance from the reference sample within a block increases, prediction accuracy decreases. Therefore, after dividing the block into line units and performing encoding and decoding, encoding of the next line can be performed.

When dividing an NxN block by line, it can be divided into N 1xN blocks or N Nx1 blocks. The following figure shows an example related to this. In this figure, the block is assumed to be 4x4 in size, and is assumed to be vertically divided into four 1x4 blocks. Additionally, it was assumed that the intra-screen prediction mode used was mode 18 in Figure 1.

Figure 5

As shown in (a) of the figure above, predicted samples P11 to P41 are generated using reference samples A` to D`. Afterwards, residual values corresponding to P11 to P41 are generated and then encoded and decoded to generate restored samples R11 to R41. Then, as in (b), R11 to R41 are set as reference samples to generate prediction samples P12 to 42. Afterwards, residual values corresponding to P12 to P42 are generated, and then encoding and decoding are performed to generate restored samples R12 to R42. When encoding is performed in this way, the distance between the reference sample and the prediction sample is always 1.

The intra-screen prediction modes that can be used in the example above are the same as Figure 1, but the available intra-screen prediction modes can be adjusted depending on the horizontal and vertical ratio of the block. For example, if the vertical length of the divided block is longer than the horizontal length as shown in Figure 5 above, only the on-screen prediction mode pointing to the left of the block (for example, modes 2 to 34 in Figure 1) is available. It is also possible.

Alternatively, it is also possible to use a different prediction mode within the screen for each line used. Alternatively, the intra-screen prediction mode can be used by adding a random offset to the intra-screen prediction mode used in the previous line. The following figure shows an example where the on-screen prediction mode used in the first line is 18 and the offset is -1.

Figure 6

The offset used in Figure 6 above can be signaled for each block.

Or, unlike Figure 6 above, rather than performing encoding and decoding for each line and moving on to the next line, only the prediction mode within the screen is set differently for each line, and all predicted values are generated in block units, and then encoded and decoded in block units. It is also possible to perform . The following figure shows an example related to this.

Figure 7

Alternatively, rather than signaling the intra-screen prediction mode on a block basis and additionally signaling the offset used for each line, it is possible to assign to a line using only the intra-screen prediction mode that exists in the MPM list. In this case, MPM index information may be signaled for each line.

Alternatively, unlike the example above, a weighted sum can be performed by using two intra-screen prediction modes for each block and applying different weights to each line. The following figure shows an example of the matrix to be used when applying a weighted sum.

Figure 8

For example, if the two modes used in the figure above are the planar mode and mode 18 in Figure 1, a prediction sample is generated for the corresponding line using the two modes, and then the final prediction sample is generated by weighted sum. In the figure above, the predicted samples corresponding to P11 to P41 are the result of a 7:1 weighted sum of the samples generated by the planar mode and mode 18. Similarly, the predicted samples corresponding to P14 to P44 are the result of a 1:7 weighted sum of the samples generated by the planar mode and the 18 mode.

Alternatively, it is also possible to create a weighted reference sample line and use it for intra-screen prediction. The following picture shows an example related to this.

Figure 9

As shown in the figure above, when performing intra-screen prediction using mode 16 for the third line, rather than using R12 to R42 as reference samples for the third line, the weighted sum of R12` to R42` is used. You can use it. R12' to R42' can be generated as a weighted sum of the reference samples existing in the first line and the reference samples existing in the second line, depending on the prediction direction.

When encoding and decoding are performed on a block basis rather than on a line basis, conversion may be performed on the residual value. At this time, 1-D conversion may be performed on a line basis, or 2-D conversion may be performed on a block basis. At this time, it can be signaled whether 1-D or 2-D transformation was used.

Claims (1)

Video signal encoding/decoding method using intra prediction.