| With the widespread adoption of internet technology and mobile devices,individuals can now watch and share videos from anywhere and at any time through the network.Moreover,4K and 8K videos offer viewers a more delicate visual experience.However,with that came an explosion of video data.In order to cope with the expanding requirements for storing and transmitting video content,it is urgent to develop video coding standards with higher compression efficiency.In this context,a new generation of Versatile Video Coding(VVC)came into being.When compared to the previous generation video coding standard HEVC(High Efficiency Video Coding),the latest video coding standard VVC(Versatile Video Coding)can achieve a nearly 50%reduction in bitrate while maintaining the same video quality.On the basis of the original technology,VVC has improved and added some coding tools,improving the coding performance while increasing the algorithm complexity sharply.Therefore,this paper conducts an optimization study on the bitrate control and quantification in the VVC coding standard.Aiming at the bitrate control technology,this paper uses the bitrate control optimization algorithm based on Transform Coefficient Distribution(TCD)to improve the Rate-Distortion(RD)performance.For quantization,this paper designs a fast quantization algorithm based on context adaptive threshold pruning to reduce the coding complexity.The specific introduction is as follows:(1)VVC adopts a variety of advanced coding tools,compared with HEVC,there is less redundancy after prediction,which is manifested in the TCD being more concentrated near the zero coefficient.Firstly,this paper explores the distortion estimation model based on the transformation coefficient distribution,and suggests a frame-level distortion estimation algorithm that uses a generalized Gaussian distribution to model the probability density function(PDF)of frame-level total coding distortion(TCD).,and represents the frame-level coding distortion as a function of distribution parameters and quantization parameters.Secondly,under the target bit constraint of the GOP(Group of Picture)level,the proposed distortion estimation model is used to analyze the distortion dependence between frames,and the optimized frame-level bit allocation is realized by combining the frame-level distortion dependence factor.Experiments show that this method can effectively improve the distortion performance of the coding rate,and the average distortion performance improvement of 1.06% and 0.08% is obtained in the low-delay P-frame(LDP)and low-delay B frame(LDB)configurations.(2)The quantization module of VVC incorporates a dependent scalar quantization algorithm(DQ)based on dynamic programming,which greatly enhances the coding performance.However,this also results in a sharp increase in computational complexity.The high coefficient dependency and complex quantization candidate search space lead to low efficiency and distortion of traversal calculation rate.This paper proposes a fast DQ algorithm based on context adaptive threshold pruning,which greatly reduces the dynamic programming search space and solves the problem of high complexity of quantitative candidate search.Through the principle of DQ algorithm and statistical analysis of quantification results,it is found that DQ quantification results are closely related to context variables such as quantification remainder,coefficient position,and neighborhood quantification results.Firstly,the complex dynamic programming quantification is abstracted into a multivariate and multi-interval classification problem of remainder,location,and neighborhood quantification results,and aiming at the problem that there are still different quantitative results in the same interval,a pruning method based on threshold comparison is proposed by analyzing the cumulative distribution function of samples in the same interval,and some "safe" quantification candidates are clipped,the search space is reduced,and the full-path search is simplified.Experimental results show that the proposed fast quantization algorithm has an average loss of rate distortion performance was 0.19% and 0.34%,and the average reduction of coding complexity was4.31% and 3.36%. |
