基于多恰可感知差值等级的高效视频编码区域级量化参数级联策略研究

高效的量化参数级联策略可以显著提高视频编码的性能，有效保证视频通信系统的运行，具有重要的理论研究意义。本项目针对目前主流的高效视频编码标准HEVC，将有效衡量视频感知质量的多恰可感知差值等级应用于量化参数级联策略研究中，提出有效提高编码感知率失真性能的区域级量化参数级联策略。项目的创新性研究包括：综合考虑视频区域的内容特性、时域抖动效应、人眼选择注意机制的影响，建立衡量视频区域感知质量的多恰可察觉差值等级与量化参数的关系模型；进一步借鉴使用lambda域的码率模型，分析视频区域多恰可察觉差值等级及码率的空时域依赖关系；基于拉格朗日率失真优化理论，构建区域联合优化的率失真代价函数并提出区域级量化参数级联策略；基于率失真代价预测的思路提出快速量化参数级联策略。相关研究成果亦可用于视频质量评价、码率控制，是优化视频编码技术，保证视频通信系统高效运行的重要理论基础。

Efficient Quantization parameter cascading (QPC) strategy improves video coding performance significantly. Also, it ensures the efficient operation of the video communication systems and has important theoretical meaning. This project focuses on the High Efficient Video Coding (HEVC) which is a mainstream video coding standard and applies an efficient perceptual video quality metric, i.e., multiple just noticeable difference (JND) levels, to the QPC strategy development. This project aims to propose regional QPC strategies for effectively improving the perceptual rate distortion (RD) performance of video coding. The innovation works include: considering the influence of the regional video content property, the temporal pumping artifacts, and the selective mechanism of human eyes, the relationship between the quantization parameter (QP) and the multiple JND levels is developed to evaluate the regional perceptual video quality. Further, the lambda domain bitrate model is utilized to analyze the regional temporal dependency relationships of the regional multiple JND levels and the bitrate, respectively. Then the regional joint optimization RD cost function and the related regional QPC strategies are developed based on the Lagrange RD optimization theory. Finally, fast QPC strategies are analyzed based on the RD cost estimation spirit. Related results can be utilized in video quality assessment and rate control. Also, the study in this project provides important theoretical basis for optimizing of the video coding techniques and the efficient operation of the video communication systems.