CN107580224B - An Adaptive Scanning Method for HEVC Entropy Coding - Google Patents

Abstract

The invention discloses a kind of adaptive scanning method towards HEVC entropy coding, specifically: adaptive scanning and entropy coding are carried out to the relevant syntactic element in the last one nonzero coefficient position in TU first；Then the syntactic element sig_coeff_flag in the syntactic element CSBF of each sub-block in TU and sub-block in each coefficient positions is determined；It calculates separately CSBF value and sig_coeff_flag value is equal to the prediction probability of minimum probability symbol LPS, and be ranked up, scan, encode respectively；Finally to the adaptive scanning and entropy coding of nonzero coefficient amplitude relevant syntax elements all in TU.Adaptive scanning method of the present invention further promotes code efficiency so that intersymbol correlation is adequately utilized by HEVC；Without transmitting additional code stream information for decoding end more；And there is lower encoder complexity, the 0.33% of only existing algorithm.

Description

An Adaptive Scanning Method for HEVC Entropy Coding

technical field

The invention belongs to the technical field of video coding, and relates to an adaptive scanning method oriented to HEVC entropy coding.

Background technique

Video can bring more vivid subjective feelings to users, and it has gradually penetrated into every corner of our lives. With the emergence of high-definition/ultra-high-definition video such as "high resolution, high frame rate, and high pixel bit depth", more and more users participate in related video services and experience the experience of high-definition/ultra-high-definition video immersively. unique charm. However, the huge amount of video data not only increases the burden on video storage systems and global communication networks, but also affects the deployment and practical application of new services. Therefore, the new-generation video coding standard HEVC (High Efficiency Video Coding, High Efficiency Video Coding) is based on predictive coding, transform coding, and entropy coding. bandwidth or transmission bandwidth.

HEVC still uses the traditional block-based hybrid coding structure, mainly including several modules such as intra/inter prediction, transformation, quantization, entropy coding, and deblocking filtering. Compared with previous coding standards, HEVC also introduces some new coding technologies: block representation based on quadtree structure, unified directional intra-frame prediction, advanced motion vector prediction, large-size discrete cosine transform, adaptive Scanning, context-adaptive binary arithmetic coding with better performance, sample-point adaptive compensation filtering, etc.

In describing the prediction residual, HEVC adopts the TU (Transform unit) mode based on the quadtree partition structure. The optional sizes of TU are 4×4, 8×8, 16×16, and 32×32. After adaptive scanning technology, a two-dimensional prediction residual in TU is mapped to a one-dimensional array to realize subsequent entropy coding operation. Among them, adaptive scanning technology, as a key step in HEVC, directly affects the final encoding bit rate. According to the principle of rate-distortion optimization, under the premise of given video reconstruction quality, the final encoding bits of one-dimensional coefficients obtained by adaptive scanning technology should be kept to the minimum.

Zigzag scanning technology is widely used in video coding, and the residual coefficients in intra/inter blocks are mapped into one-dimensional arrays in a fixed scanning order (see Wiegand T., Sullivan G., Bjontegaard G., Luthra A. .Overview of the H.264/AVC video coding standard[J].IEEE Trans.Circuits Syst.Video Technol.,2003). Fouak et al. combined Zigzag scanning, horizontal scanning, vertical scanning and Hibert scanning to establish a new scanning method (Fouak F., Benzid R., Benoudjit N.. Color image compression algorithm based on the DCT transform combined to an adaptive block scanning[J].Int.J.Electron.Commun.,2011). Since diagonal scanning can better reduce data redundancy information, Sole et al. proposed to replace Zigzag scanning with diagonal scanning, combine horizontal scanning and vertical scanning, and obtain a new adaptive scanning method, and apply this method to the In HEVC coding (Sole J., Joshi R., Nguyen N.. Transform coefficient coding in HEVC [J]. IEEETrans. Circuits Syst. Video Technol., 2012). The above method takes into account the distribution characteristics of the prediction parameter coefficients, which improves the coding efficiency to a certain extent. However, since the scanning model is relatively fixed, the characteristics of subsequent entropy coding are ignored, which limits the further improvement of coding performance. Therefore, it is necessary to establish an adaptive scanning method suitable for HEVC entropy coding when coding the prediction residual of the TU layer.

Contents of the invention

The purpose of the present invention is to provide an adaptive scanning method oriented to HEVC entropy coding, which can improve the coding performance of HEVC and has low coding complexity.

The technical scheme adopted in the present invention is a kind of adaptive scanning method for HEVC entropy coding, specifically implemented according to the following steps:

Step 1, performing adaptive scanning and entropy coding on the syntax element related to the position of the last non-zero coefficient in the TU;

Step 2, perform adaptive scanning and entropy coding on the syntax elements related to the positions of other non-zero coefficients in the TU:

2.1 Divide a TU into one or more 4×4 sub-blocks; determine the syntax element CSBF of each sub-block and the syntax element sig_coeff_flag at each coefficient position in the sub-block according to the quantized residual coefficient; CSBF is used to identify whether each sub-block contains non-zero coefficients, and sig_coeff_flag is used to identify the position of non-zero coefficients in the sub-block;

2.2 Calculate the CSBF value of each syntax element and the sig_coeff_flag value of each syntax element equal to the predicted probability of the minimum probability symbol LPS, and perform sorting, scanning, and encoding respectively;

Step 3, adaptive scanning and entropy coding of all syntax elements related to the amplitude of non-zero coefficients in the TU.

The present invention is also characterized in that,

Step 1 is specifically: use the original adaptive scanning method in HEVC to obtain the coordinates (X, Y) of the last non-zero coefficient of the prediction residual in the TU; according to the coordinate value, obtain the syntax related to the position of the last non-zero coefficient Elements: last_sig_coeff_x_prefix, last_sig_coeff_y_prefix, last_sig_coeff_x_suffix, last_sig_coeff_y_suffix, and scan and entropy encode the four syntax elements in turn.

Step 2.2 is specifically:

① Obtain the context probability index and LPS corresponding to each CSBF and sig_coeff_flag respectively;

② Predict the probability that each CSBF value and sig_coeff_flag value is equal to its corresponding LPS;

③ Regionally divide CSBF and sig_coeff_flag respectively, and classify CSBF and sig_coeff_flag with the same context probability index value into the corresponding same sub-region;

④ Sort the sub-regions corresponding to CSBF and sig_coeff_flag respectively, perform scanning and entropy coding.

The probability when predicting the corresponding LPS is specifically: the definition symbol s represents the syntax element, and S represents the encoded syntax element sequence; according to the KT estimation method, the probability p 1 of s= ₁ and the probability p ₀ of s=0 are respectively Yes:

p ₀ =1-p ₁

Among them, n ₀ and n ₁ respectively denote the occurrence frequency of the value 0 and the value 1 in the syntax element sequence S.

The sub-region sorting requirements are as follows: the sub-regions are sorted according to the corresponding context probability index values from large to small, and the sub-regions are sorted according to the corresponding predicted probability values from small to large.

Step 3 is specifically: use the original adaptive scanning method in HEVC to obtain the syntax elements corresponding to each non-zero coefficient in the TU: coeff_abs_level_greater1_flag, coeff_abs_level_greater2_flag, coeff_sign_flag, and coeff_abs_level_remaining, and scan and entropy encode the four syntax elements in turn.

The beneficial effect of the present invention is,

(1) The present invention predicts the occurrence probability of the syntax element value to be encoded from the perspective of probability theory, and lays a theoretical foundation for the establishment of the HEVC adaptive scanning method.

(2) The present invention considers the operation mechanism of entropy coding to establish an adaptive scanning method, so that the correlation between symbols is fully utilized by HEVC, and the coding efficiency is further improved.

(3) The present invention utilizes known encoding information to establish an adaptive scanning method, without needing to transmit additional code stream information to the decoding end.

(4) The adaptive scanning method proposed by the present invention has a relatively low encoding complexity, which is only 0.33% of the existing algorithm.

Description of drawings

Fig. 1 is a flow chart of an adaptive scanning method for HEVC entropy coding of the present invention;

Fig. 2 is a comparison diagram of the rate-distortion performance between the method of the present invention and the existing algorithm of HEVC under the test sequence of BQMall;

Fig. 3 is a comparison chart of the rate-distortion performance between the method of the present invention and the existing algorithm of HEVC under the test sequence of FourPeople.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments.

The present invention provides a flow chart of an adaptive scanning method oriented to HEVC entropy coding as shown in Figure 1, specifically implemented according to the following steps:

Step 1, according to the existing adaptive scanning method of HEVC, obtain the two-dimensional coordinates of the last non-zero coefficient position in the TU;

Step 2, according to the two-digit coordinates in step 1, determine the syntax elements last_sig_coeff_x_prefix, last_sig_coeff_y_prefix, last_sig_coeff_x_suffix, last_sig_coeff_y_suffix values related to the position of the last non-zero coefficient, and then perform entropy coding;

Step 3, obtain the syntax element CSBF related to the position of each non-zero coefficient in the TU;

Step 4, obtain the context probability index and LPS corresponding to each CSBF;

Step 5, predict the probability when each CSBF value is equal to its corresponding LPS according to formula (1);

Among them, s represents a syntax element, S represents a coded syntax element sequence, p ₁ is the probability of s=1 and p ₀ is the probability of s=0, n ₀ and n ₁ represent the values of 0 and 0 in the syntax element sequence S respectively The frequency of occurrences with a value of 1

Step 6, divide the CSBF into regions, and group the CSBFs with the same context probability index value into a subregion;

Step 7: The sub-regions are sorted according to the context probability index values corresponding to CSBF from large to small, and the sub-regions are sorted according to the predicted probability values corresponding to CSBF from small to large. Scanning and entropy encoding are performed according to the arranged CBFS sequence;

Step 8, obtain the syntax element sig_coeff_flag related to the position of each non-zero coefficient in the TU;

Step 9, obtain the context probability index and LPS corresponding to sig_coeff_flag;

Step 10, predicting the probability that each sig_coeff_flag value is equal to its corresponding LPS according to formula (1);

Step 11, divide sig_coeff_flag into regions, and group sig_coeff_flags with the same context probability index value into a subregion;

Step 12, sort the sub-regions according to the context probability index values corresponding to sig_coeff_flag from large to small, and sort within the sub-regions according to the predicted probability values corresponding to sig_coeff_flag from small to large. Perform scanning and entropy encoding according to the arranged sig_coeff_flag sequence;

Step 13, obtain the syntax elements coeff_abs_level_greater1_flag, coeff_abs_level_greater2_flag, coeff_sign_flag, and coeff_abs_level_remaining related to the magnitude of the non-zero coefficient, and perform scanning and entropy coding on these four syntax elements in sequence.

In the self-adaptive scanning of syntax elements related to the position of the last non-zero coefficient in the TU, on the one hand, the present invention considers that the number of syntax elements involved in the position of the last non-zero coefficient is small, and on the other hand, it ensures that the proposed technology does not affect HEVC decoding Therefore, the original adaptive scanning method in HEVC is used for scanning.

The non-zero coefficient position information in TU is encoded by the constant path model in entropy coding. The adaptive scanning method adopted by HEVC directly affects the final coding efficiency. Therefore, the non-zero coefficient position correlation is established by considering the operation mode of entropy coding. An adaptive scanning method for syntax elements is necessary. The specific method is to first divide a TU into one or more 4×4 sub-blocks; according to the quantized residual coefficient, determine the syntax element CSBF (coded_sub_block_flag) of each sub-block and the position of each coefficient in the sub-block The syntax element sig_coeff_flag on. CSBF is used to identify whether each sub-block contains non-zero coefficients, and sig_coeff_flag is used to identify the position of non-zero coefficients in the sub-block.

Secondly, the symbol s is defined to represent a syntax element, and S represents an encoded sequence of syntax elements. According to the KT estimation method, the probability p 1 of s= ₁ and the probability p ₀ of s=0 are respectively:

p ₀ =1-p ₁

Among them, n ₀ and n ₁ respectively denote the occurrence frequency of the value 0 and the value 1 in the syntax element sequence S. Calculate the predicted probability that the CSBF value of each syntax element is equal to the minimum probability symbol (LPS), and calculate the predicted probability that the value of each syntax element sig_coeff_flag is equal to the LPS.

According to the principle of entropy coding, the context probability index used by the syntax element CSBF is acquired. The CSBF syntax elements with the same context probability index number are grouped into a sub-area, and the sub-areas are sorted from large to small according to the value of the context probability index number, and the predicted probability value of each syntax element CSBF in the sub-area is from small to large Sorting is performed, and the sorted CSBFs are scanned sequentially and sent to the entropy encoder.

According to the principle of entropy coding, the context probability index used by the syntax element sig_coeff_flag is acquired. The sig_coeff_flag syntax elements with the same context probability index number are grouped into a sub-region, and the sub-regions are sorted according to the context probability index number value from large to small, and the predicted probability value of each syntax element sig_coeff_flag in the sub-region is from small to large Sorting is performed, and the sorted sig_coeff_flags are scanned sequentially and sent to the entropy encoder.

In TU, the non-zero coefficient amplitude is mainly encoded by the bypass model in entropy coding. The adaptive scanning method adopted by HEVC has little influence on the final number of encoded bits, so the original adaptive scanning method in HEVC is used for scanning. . Use the original adaptive scanning method in HEVC to obtain the syntax elements corresponding to each non-zero coefficient in the TU: coeff_abs_level_greater1_flag, coeff_abs_level_greater2_flag, coeff_sign_flag, and coeff_abs_level_remaining, and scan and entropy encode these four syntax elements in turn.

The effects of the present invention are further illustrated by experiments.

The experiments tested the encoding performance of the proposed method in different resolutions and different application scenarios. FIG. 2 shows a comparison chart of the rate-distortion performance between the present invention and the original HEVC algorithm under the test sequence of Cactus. Fig. 3 shows a comparison chart of the rate-distortion performance between the present invention and the original HEVC algorithm under the test sequence of BQMall. It can be seen from FIGS. 2 and 3 that the HEVC entropy coding-oriented adaptive scanning method proposed in the present invention has higher coding efficiency than the original HEVC scanning method. Statistical results show that the peak signal-to-noise ratio PSNR of video quality obtained by the present invention is 0.09dB higher on average than the PSNR of video quality obtained by the original method of HEVC.

Claims (1)

1. a kind of adaptive scanning method towards HEVC entropy coding, which is characterized in that specifically implement according to the following steps:

Step 1, adaptive scanning and entropy coding are carried out to the relevant syntactic element in the last one nonzero coefficient position in TU, specifically Are as follows: using original adaptive scanning method in HEVC, obtain prediction residual the last one nonzero coefficient in TU coordinate (X, Y)；According to the coordinate value, the last one relevant syntactic element in nonzero coefficient position: last_sig_coeff_x_ is obtained Prefix, last_sig_coeff_y_prefix, last_sig_coeff_x_suffix, last_sig_coeff_y_ Suffix, and four syntactic elements are successively scanned and entropy coding；

Step 2, adaptive scanning and entropy coding are carried out to the relevant syntactic element in other nonzero coefficient positions in TU:

2.1 are divided into a TU sub-block of one or more 4 × 4 sizes；According to the residual error coefficient after quantization, each is determined Syntactic element sig_coeff_flag in the syntactic element CSBF and sub-block of sub-block in each coefficient positions；CSBF is used for Identify whether each sub-block contains nonzero coefficient, sig_coeff_flag is used to identify the position of nonzero coefficient in sub-block；

2.2 calculate each syntactic element CSBF value and each syntactic element sig_coeff_flag value equal to minimum probability The prediction probability of symbol LPS, and be ranked up, scan, encode respectively, specifically:

1. obtaining the corresponding context probability index of each CSBF and sig_coeff_flag and LPS respectively；

2. predict probability when each CSBF value and sig_coeff_flag value are equal to the LPS corresponding to it, specifically: definition Symbol s indicates that syntactic element, S indicate encoded syntactic element sequence；According to K-T estimation method, then the Probability p of s=1₁And s =0 Probability p₀It is respectively:

p₀=1-p₁

Wherein n₀And n₁Respectively indicate syntactic element sequence S intermediate value be 0 and value be 1 appearance the frequency；

3. region division is carried out to CSBF and sig_coeff_flag respectively, respectively by context probability index value having the same CSBF and sig_coeff_flag be grouped into corresponding same subregion；

4. being ranked up respectively to the corresponding subregion of CSBF and sig_coeff_flag, it is scanned and entropy coding；

Subregion ordering requirements are as follows: be ranked up between subregion according to corresponding context probability index value is descending, sub-district It is ranked up in domain according to corresponding prediction probability value is ascending；

Step 3, to the adaptive scanning and entropy coding of nonzero coefficient amplitude relevant syntax elements all in TU, specifically: it utilizes Original adaptive scanning method in HEVC obtains the corresponding syntactic element of each nonzero coefficient in TU: coeff_abs_ Level_greater1_flag, coeff_abs_level_greater2_flag, coeff_sign_flag and coeff_abs_ Level_remaining, and four syntactic elements are successively scanned and entropy coding.