KR101845622B1 - Adaptive rdpcm method for video coding, video encoding method based on adaptive rdpcm and video decoding method based on adaptive rdpcm - Google Patents

Abstract

The decoding method based on the adaptive RDPCM includes a step of extracting information on whether the decoder performs the adaptive RDPCM in the encoded bitstream, the step of decoding the current block, which is intended to perform RDPCM, on the decoded current frame based on the inter- Determining a weight for a sample block that is adjacent to the current block and for which a residual signal is predicted by RDPCM; and determining, based on the result of inter-picture prediction for the current block, a residual signal for the sample block, And performing RDPCM on the current block to decode the residual signal.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive RDPCM method, an encoding method based on adaptive RDPCM, and a decoding method based on adaptive RDPCM,

The technique described below relates to a video coding technique based on RDPCM.

Recently, with the development of computer synthesis technology and digital contents production technology, screen contents video produced by computer synthesis is widely used for new multimedia service. Screen content video contains many graphics, characters, and numbers with high contrast ratio, unlike ordinary video. Therefore, existing compression technologies such as HEVC (High Efficiency Video Coding) video compression standard developed for HD / UHD broadcasting Efficient compression may not be easy.

In the HEVC extension standard, a scheme of performing residual differential pulse-code modulation (RDPCM) on the residual signal when encoding a transform unit (TU) in a transform skipping mode in lossy compression has been developed. Conversion skip mode is a technique of direct entropy coding without transforming the residual signal, which is useful for screen content video compression because there are many residual components in the high frequency domain that occur at the border of graphical elements with high contrast. RDPCM reduces the total energy of residual components for entropy encoding in the process of omitting the conversion, thereby providing better compression performance.

Y. L. Lee, K.H. Han, and G. Sullivan, "Improved lossless intra coding for H.264 / AVC," IEEE Trans. Image Processing, 2006. M. Nacarri, A. Gabriellini, M. Mrak, S. Blasi and E. Izquierdo, "Inter prediction Residual DPCM ", JCTVC-M0442, Apr. 2013.

The technique described below is intended to provide an RDPCM technique based on adjacent sample residual signals.

The adaptive RDPCM method for an image comprises the steps of: determining a plurality of sample blocks adjacent to a current block for which RDPCM prediction is to be performed in a frame in which inter-picture prediction is performed, in which RDPCM prediction is performed; Determining a weight for a quantized residual signal in the plurality of sample blocks based on an area in which RDPCM prediction is performed, and a step of calculating, using the weighted value of the quantized residual signal, And performing RDPCM on the block.

An encoding method based on an adaptive RDPCM includes a step of performing an inter-picture prediction on an current frame by an encoder, a step of performing an inter-picture prediction on a current block by using an RDPCM The encoder performs RDPCM on the current block using the result of the inter-picture prediction for the current block, the residual signal for the sample block, and the weight, and predicts the residual signal And entropy coding the residual signal for the current block by the encoder.

The decoding method based on the adaptive RDPCM includes a step of extracting information on whether the decoder performs the adaptive RDPCM in the encoded bitstream, the step of decoding the current block, which is intended to perform RDPCM, on the decoded current frame based on the inter- Determining a weight for a sample block that is adjacent to the current block and for which a residual signal is predicted by RDPCM; and determining, based on the result of inter-picture prediction for the current block, a residual signal for the sample block, And performing RDPCM on the current block to decode the residual signal.

The technique described below can predict the current residual signal effectively by assigning the learned weight to the sample residual signal adjacent to the residual signal to be predicted.

1 is an example of a process of performing V-RDPCM.
2 is an example showing the position of a sample for residual signal prediction.
3 is an example of an area for performing adaptive V-RDPCM.
Figure 4 is an example of a flowchart for an image encoding method based on adaptive RDPCM.
5 is an example of a flowchart for an image decoding method based on adaptive RDPCM.

The following description is intended to illustrate and describe specific embodiments in the drawings, since various changes may be made and the embodiments may have various embodiments. However, it should be understood that the following description does not limit the specific embodiments, but includes all changes, equivalents, and alternatives falling within the spirit and scope of the following description.

The terms first, second, A, B, etc., may be used to describe various components, but the components are not limited by the terms, but may be used to distinguish one component from another . For example, without departing from the scope of the following description, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

As used herein, the singular " include "should be understood to include a plurality of representations unless the context clearly dictates otherwise, and the terms" comprises & , Parts or combinations thereof, and does not preclude the presence or addition of one or more other features, integers, steps, components, components, or combinations thereof.

Before describing the drawings in detail, it is to be clarified that the division of constituent parts in this specification is merely a division by main functions of each constituent part. That is, two or more constituent parts to be described below may be combined into one constituent part, or one constituent part may be divided into two or more functions according to functions that are more subdivided. In addition, each of the constituent units described below may additionally perform some or all of the functions of other constituent units in addition to the main functions of the constituent units themselves, and that some of the main functions, And may be carried out in a dedicated manner.

Also, in performing a method or an operation method, each of the processes constituting the method may take place differently from the stated order unless clearly specified in the context. That is, each process may occur in the same order as described, may be performed substantially concurrently, or may be performed in the opposite order.

RDPCM of screen content coding (HEVC / SCC) is an implicit RDPCM method that performs RDPCM in the same direction after intra-picture prediction in the horizontal and vertical directions, and an implicit RDPCM method in which a horizontal (H- An explicit RDPCM scheme may be used in which the RDPCM is selectively performed and the RDPCM prediction direction information is transmitted to the decoding side through a bitstream. The following description relates to an explicit RDPCM scheme. The coding described below is performed in the encoder and decoder. On the other hand, the encoder and the decoder may be referred to as a computer apparatus or an image processing apparatus which processes the image constantly.

1 is an example of a process of performing V-RDPCM. 1 (a) is a process of performing V-RDPCM in lossless coding, and FIG. 1 (b) is an example of a process of performing V (Vertical) -RDPCM in lossy coding. V-RDPCM performs RDPCM in the vertical direction. In FIG. 1 (a), r _{i, j} means residual samples in N × N blocks. The current residual sample r _{i, j} is predicted from the upper residual sample r _{i -1, j} . For reference, the current residual sample r _{i, j} in H-RDPCM is predicted from the left residual sample r _{i, j-1} . The lossy coding of FIG. 1 (b) differs from the lossless coding of FIG. 1 (a) in that each residual sample is predicted from a quantized residue.

는 아래의 수학식 1로 표현할 수 있다. The residual residual signal after RDPCM application in the vertical direction with respect to the residual signal r _{i, j} of the NxN block in lossy coding

Can be expressed by the following equation (1).

를 엔트로피 부호화 후 디코더에 전송하고 다음 행의 잔차 신호 예측을 위하여 복원하여 예비한다. 해당 예측 과정은 블록의 모든 행에 대해서 순차적으로 진행한다. 반대로 디코더는 아래의 수학식 2와 같이 복원한 2차 잔차 신호를 순차적으로 더하여 현재 i번째 행의 잔차 신호를 다음과 같이 복원한다.Q (r) is the reconstructed residual signal including the quantization noise. In the case of V-RDPCM,

Is entropy-coded and then transmitted to the decoder, and is restored for prediction of the residual signal of the next row. The prediction process proceeds sequentially for all the rows of the block. Conversely, the decoder sequentially adds the restored second residual signals as shown in the following Equation 2 to restore the residual signal of the current i-th row as follows.

The coding scheme described below relates to RDPCM performed in the process of performing lossy compression according to the HEVC extension standard. The RDPCM described below performs prediction using adjacent residual components for the current residual samples. The RDPCM described below performs a more accurate prediction from the weighted sum of adjacent residual components for the current residual samples. The RDPCM described below is called an adaptive RDPCM. The encoder follows an explicit RDPCM scheme that determines whether to perform adaptive RDPCM through bit rate-distortion optimization and transmits it to the bit stream. The adaptive RDPCM performs prediction on residual samples as shown in Equation (3) below. Hereinafter, an N × N size block will be described as a reference. For convenience of explanation, the adaptive V-RDPCM will be mainly described.

는 잔차 신호 r_i,j의 주변에 위치하면서 r_i,j의 예측에 사용되는 양자화된 잔차 신호이다. In Equation (3), K denotes a neighboring region that has already been encoded, and k ₁ and k ₂ are position indices of residual samples used for prediction of the residual signal r _{i, j} performing the current encoding. alpha is a weight parameter.

_J is a quantized residual signal used for prediction of r _{i, j} , located around the residual signal r _{i, j} .

The encoder can determine α by learning based on a sample that has already been predicted. In this case, the encoder does not transmit? As a bitstream to the decoder, but the decoder can learn? Based on the already decoded sample.

Samples for prediction of the current residual signal can be determined in consideration of coding performance, complexity, and parallel processing. 2 is an example showing the position of a sample for residual signal prediction. 2 (a) is an example of a general RDPCM, FIG. 2 (b) is an example of an adaptive V (vertical) -RDPCM and FIG. 2 (c) is an example of an adaptive H (Horizontal) .

As shown in FIG. 2 (a), the left and upper samples immediately adjacent to the residual signal r _{i, j} to be currently encoded can be used. Also, using two directions, left and top, may be difficult to parallelize. Therefore, considering the parallel processing, only the upper sample can be used for the residual signal r _{i, j} in case of V-RDPCM as shown in FIG. 2 (b). In the case of H-RDPCM, only the left sample can be used for the residual signal r _{i, j as shown} in FIG. 2 (c). Furthermore, considering the complexity of the calculation, only some samples can be used without using all of the samples located on the left side of the residual signal r _{i, j} or all the samples located on the upper side.

In the case of the V-RDPCM, the adaptive RDPCM described below has the upper left side (r _{i-1, j-1} ) samples at the position of the residual signal r _{i, j} to be predicted as shown in FIG. 2 (r _{i-1, j 1} ) sample and an upper right side (r _{i-1, j + 1} ) sample. In this case, the residual signal prediction for the V-RDPCM in the adaptive RDPCM is expressed by Equation (4) below. In Equation (4), j is 0 to N-1.

The H-RDPCM Figs. 2 (c) residual signal, such as r _i, the upper left side _{(r i-1, j-} 1) of the left sample at the position of _j samples, the left sample (r _{i, j-1)} and the left lower Side (r _{i + 1, j-1} ) samples can be used.

We now have to determine the weights for the samples to perform the prediction on the residual signal r _{i, j} . The encoder may encode the weight a for each sample in Equation (4) into a bitstream and transmit it. However, if a plurality of weights used by the encoder for all residual signals are transmitted to the decoder, there is a problem that the amount of data to be transmitted increases. Therefore, basically, the encoder determines the weight for each sample used for prediction of the residual signal r _{i, j} through learning. The encoder learns the weights of neighboring samples for the residual signal r _{i, j} using the region where the residual signal has already been learned.

3 is an example of an area for performing adaptive V-RDPCM. In Fig. 3, Ω _{i, j} represents an area for learning a weight of a sample for the current residual signal r _{i, j} . In Fig. 3, Ω _{i, j} is shaded with M × 3 size blocks. In Fig. 3, Ω _{i, j} includes three first upper samples adjacent to the residual signal r _{i, j} and three second upper samples adjacent to the upper side of the first upper sample. Ω _{i, j} = {r _{im, jn} | m = 1, ..., M, and n = -1, 0, 1}. On the other hand, Ω _{i, j} can use a region different from that of FIG. However, Ω _{i, j} should basically be a region that has already been predicted. It is also desirable to include samples immediately adjacent to the residual signal r _{i, j} to be predicted. For convenience of explanation, it is assumed that M = 2.

In FIG. 3,? ^L _{i, j} denotes any one of the samples located in the region? _{I, j} . l = 1, ..., | Oi _{, j} |. The distortion value D _{i, j} for? ^L _{i, j} can be defined as Equation (5) below.

는 아래의 수학식 6과 같다.In Equation (5), d (l) is the Manhattan distance between r _{i, j} and Ω ^l _{i, j} . The closer d (l) is to r _{i, j} , the higher the weight. σ _d can be 2.5, which can be determined experimentally. The second-order residual signal at? _{I, j}

Is expressed by Equation (6) below.

In Equation (6), K means the three predetermined sample positions at? ^L _{i, j} for the current residual signal r _{i, j} . α _{k1 and k2} are model coefficients. The model coefficients? _{K1 and k2} can be determined to minimize the above Equation (5). The model coefficients? _{K1, k2} minimizing Equation (5) can be determined in various ways. For example, techniques such as least square optimization may be used. The least square method is vulnerable to the approximation of outliers, and there is a problem that the interpretive power of the variable is degraded when a strong collinear relationship occurs between the sample data. In the case of the screen content, the residual signal generated at the boundary of the graphic element may not be suitable for the least squares method in the signal property of the step function type in which the variation amount between neighboring pixels in the space is very large and the boundary is small. To determine the model coefficients α _{k1 and k 2} , we can use a rare model that gives a coefficient of L ₁ penalty to the coefficient so that the estimated value of the coefficient approaches zero or approaches zero. The distortion function of Equation (5) can be modified as shown in Equation (7) below.

In Equation (6), P (?) Corresponds to the L ₁ regularization term. P (α) = Σ _{(k1, k2) ∈ K} | α _{k1, k2} | and β is a constant. Now, an optimum parameter? ^* _K that minimizes Equation (7) can be obtained as shown in Equation (8) below. ⁽ _k) of the residual signal r _{i, j} (for example,? in Equation 4) with respect to the position of the residual signal r _{i, j} .

는 가능한 α_k의 모든 집합을 의미한다. In Equation (8)

Means all possible sets of α _k .

Finally, the encoder can now perform RDPCM on the residual signal r _{i, j} . The decoder unit can perform weight determination and RDPCM on the current residual signal r _{i, j} using the decoded sample region in the same operation.

4 is an example of a flowchart for an image encoding method 100 based on adaptive RDPCM. It is assumed that the encoder starts encoding in the transcoding skip mode (110). The encoder may then determine whether to perform adaptive RDPCM (120). The encoder can determine whether to perform adaptive RDPCM based on bit rate-distortion optimization. The encoder may select one of the RDPCM off mode, the adaptive V-RDPCM mode, and the adaptive H-RDPCM mode. The RDPCM off mode is to perform encoding in the transcoding skip mode without performing RDPCM. The encoder can select the lowest cost mode based on the distortion value plus the bit rate for the three candidate modes. The encoder operates in an RDPCM off mode 130, in an adaptive V-RDPCM mode 140, or in an adaptive H-RDPCM mode, depending on the selected mode. The encoder encapsulates the information about the mode selected by the encoder into a bitstream and transmits it to the decoder.

The above description is summarized for the adaptive V-RDPCM mode or the adaptive H-RDPCM mode. It is assumed that the encoder performs explicit RDPCM. The encoder performs inter picture prediction. That is, the encoder performs prediction on a current frame using information of a previous frame, and performs prediction on a residual signal in an adaptive V-RDPCM mode or an adaptive H-RDPCM mode.

)를 예측하고자 하는 블록이 현재 프레임에서 첫 번째 행이라면 화면간 예측으로 결정된 값을 그대로 사용한다(수학식 4에서 i=0인 경우). 인코더는 잔차 신호를 예측하고자 하는 블록이 첫 번째 행이 아닌 경우 예측이 완료된 잔차 신호를 이용하여 현재의 잔차 신호에 대한 RDPCM을 수행한다. 인코더는 도 2(b)와 같이 현재의 잔차 신호를 기준으로 좌상측 샘플, 상측 샘플 및 우상측 샘플을 사용하여 현재의 잔차 신호를 예측한다. 인코더는 도 3과 같이 이미 예측이 완료된 샘플 영역을 이용하여 인접한 샘플 신호에 대한 가중치를 결정한다. 최종적으로 인코더는 인접한 샘플 신호와 가중치를 이용하여 현재의 잔차 신호를 예측한다. 인코더는 수학식 4와 같이 화면간 예측으로 예측된 값에 가중치가 부여된 인접한 샘플 신호를 감산하여 현재의 잔차 신호를 결정한다. 이와 같은 방식을 반복하여 인코더는 현재 프레임에 속한 잔차 신호에 대한 RDPCM을 수행한다. 인코더는 잔차 신호

를 엔트로피 코딩하여 디코더로 전달한다. 또한 인코더는 다음 행에 대한 잔차 신호 예측을 위하여 현재 엔트로피 코딩한 비트스트림을 복원하여 잔차 신호를 준비한다. 참고로 적응적 H-RDPCM 모드는 전술한 바와 같이 예측하는 방향만이 V-RDPCM 모드와 다르다.Based on the adaptive V-RDPCM mode, the encoder generates a residual signal (

) Is the first row in the current frame, the value determined by inter-picture prediction is used as is (i = 0 in Equation (4)). The encoder performs RDPCM on the current residual signal using the predicted residual signal if the block for which the residual signal is to be predicted is not the first row. The encoder predicts the current residual signal using the upper left sample, the upper sample, and the upper right sample based on the current residual signal as shown in FIG. 2 (b). As shown in FIG. 3, the encoder determines a weight for adjacent sample signals using a sample region that has already been predicted. Finally, the encoder predicts the current residual signal using the adjacent sample signal and the weight. The encoder determines the current residual signal by subtracting the adjacent sample signal weighted by the predicted value by the inter-picture prediction as shown in Equation (4). By repeating this method, the encoder performs RDPCM on the residual signal belonging to the current frame. The encoder generates a residual signal

And transmits the entropy-coded data to the decoder. In addition, the encoder prepares a residual signal by restoring the current entropy-coded bitstream for prediction of the residual signal for the next row. For reference, the adaptive H-RDPCM mode differs from the V-RDPCM mode only in the prediction direction as described above.

5 is an example of a flowchart for an image decoding method 200 based on adaptive RDPCM. The process performed by the decoder corresponds to the reverse order of the process performed by the encoder. The decoder first extracts information on whether to perform adaptive RDPCM in the encoded bitstream (210). If the bitstream is encoded in RDPCM off mode, general decoding is performed without performing RDPCM. If the adaptive V-RDPCM mode or the adaptive H-RDPCM mode is performed, decoding should be performed according to the corresponding mode.

The adaptive V-RDPCM mode is used as a reference. The decoder decodes the current frame based on the inter-picture prediction. If the block to be decoded is located in the first row (i = 0) of the frame, the decoder uses the current inter-picture predicted value as the residual signal as it is (Equation 4 corresponds to the case of i = 0). If the block to be decoded is located after the second row (i > 0) of the frame, the decoder first determines 220 a weight for the sample signal using at least a portion of the decoded region. The weight determination process is the same as the above-described equations (5) to (8). The decoder performs an RDPCM prediction on the current future signal by adding the sum of adjacent sample signals weighted to the current residual signal (230). The decoder performs all RDPCM prediction on the residual signal in the current frame. If the residual signal prediction for the current frame is not completed, the RDPCM prediction for the next residual signal is performed.

The present embodiment and drawings attached hereto are only a part of the technical idea included in the above-described technology, and it is easy for a person skilled in the art to easily understand the technical idea included in the description of the above- It will be appreciated that variations that may be deduced and specific embodiments are included within the scope of the foregoing description.

Claims (12)

Determining a plurality of sample blocks adjacent to a current block for which RDPCM prediction is to be performed and RDPCM prediction is performed in a frame in which inter-picture prediction is performed;
Determining a weight for a residual signal quantized in the plurality of sample blocks based on an area in which the RDPCM prediction is performed by the image processing apparatus; And
And performing the RDPCM on the current block using the weighted value of the quantized residual signal,
If RDPCM is V (vertical) -RDPCM, the plurality of sample blocks are an upper left block, an upper block, and an upper right block immediately adjacent to the current block, and when the RDPCM is H (Horizontal) The sample block is an upper left block, a left block and a lower left block immediately adjacent to the current block,
Wherein the image processing apparatus determines the weight to minimize a distortion value for the sample block in which residual prediction is completed. delete The method according to claim 1,
And if the current block is located in the first row of the frame, the residual signal for the current block uses the value calculated by inter-picture prediction as it is. The method according to claim 1,
Wherein if the current block is located after the second row in the frame, the residual signal for the current block is weighted by the weighted residual signal for each of the plurality of sample blocks in the inter- The adaptive RDPCM method for an image that is a value obtained by subtracting a value obtained by multiplying a value obtained by subtracting a value obtained by multiplying a value by Performing an inter-picture prediction on the current frame by the encoder;
Determining a weight for a plurality of sample blocks adjacent to a current block for which RDPCM is to be performed in the current frame and for which a residual signal is predicted by RDPCM;
Estimating a residual signal by performing an RDPCM on the current block using the result of the inter-picture prediction for the current block, the residual signal for the plurality of sample blocks, and the weight; And
And the encoder entropy coding the residual signal for the current block,
If RDPCM is V (vertical) -RDPCM, the plurality of sample blocks are an upper left block, an upper block, and an upper right block immediately adjacent to the current block, and when the RDPCM is H (Horizontal) The sample block is an upper left block, a left block and a lower left block immediately adjacent to the current block,
Wherein the encoder determines the weight to minimize a distortion value for the sample block. 6. The method of claim 5,
Further comprising determining whether the encoder performs the RDPCM using bit rate-distortion optimization. ≪ Desc / Clms Page number 21 > delete 6. The method of claim 5,
Wherein if the current block is located after the second row in the frame, the encoder outputs a residual signal for the current block to a quantized residual signal for each of the plurality of sample blocks from the inter- Is determined as a value obtained by subtracting a value obtained by multiplying a weight by a weight. Extracting information on whether the decoder performs adaptive RDPCM in the encoded bitstream;
Determining a weight for a plurality of sample blocks adjacent to a current block for which RDPCM is to be performed and for which a residual signal is predicted by RDPCM, in the current frame decoded based on the inter-picture prediction; And
And decoding the residual signal by performing RDPCM on the current block using the result of the inter-picture prediction for the current block, the residual signal for the plurality of sample blocks, and the weight,
If RDPCM is V (vertical) -RDPCM, the plurality of sample blocks are an upper left block, an upper block, and an upper right block immediately adjacent to the current block, and when the RDPCM is H (Horizontal) The sample block is an upper left block, a left block and a lower left block immediately adjacent to the current block,
Wherein the decoder determines the weight to minimize a distortion value for the sample block in which residual prediction is complete. delete 10. The method of claim 9,
Wherein the decoder uses the inter-picture prediction result value for the current block as the residual signal for the current block when the current block is located in the first row in the frame. 10. The method of claim 9,
Wherein when the current block is located after the second row in the frame, the decoder outputs a residual signal for the current block as a quantized residual signal for each of the plurality of sample blocks from the inter- And a value obtained by multiplying a weighted value by a weighted value is determined as an addition sum value.

Images