Research On Video Coding Algorithm Based On Visual Perception

As the rapid development of digital image technology and improvements of visual experience requirements by viewers,the research on video coding algorithm has occupied more and more important place.The traditional block-based hybrid coding scheme to improve video encoding performance is with the high coding complexity for the price,while biomedicine evolvement made another choice for the video codec technology,which based on human visual perceptual(HVS)possible,recently.HVS-based video codec technology indicates combination of human visual perception characteristics and video codec technology to achieve goals that to maximum remove the visual redundancy in video information.Obviously,HVS technology will become a trend and hotspot in the field of video codec in the future.This paper researched key technologies of visual perception which used by mainstream encoder.Achievements details are as follows:(1)The thesis gives an overview of HVS characteristics with wide application and coding method based on HVS characteristics.In practical terms,we introduce visual perception models and quality evaluation models which based on HVS characteristics proposed by scholars and give a theoretical analysis the visual perception of the rate-distortion optimization technology.This technology integrates the video encoding technology as the theoretical basis of the rate-distortion optimization technique with HVS coding method that takes advantage of video visual redundancy information to achieve a more reasonable distribution with bitrate and distortion.(2)Perceptual rate distortion optimization technology in the mainstream encoder of H.264 and HEVC were studied.In proper order we give a brief analysis on application of X264 encoder perceptual rate distortion optimization technology,and elaborate bitrate control algorithm theory.Especially we finished theoretical analysis and performance evaluation of time domain MBTree algorithm and spatial domain VAQ algorithm,which are the key technologies of visual perception and lacking of sufficient study in academia or industry.The most important we aim at empirical-based MBTree algorithm mapping model,with a number of coding parameters have the features of complex correlation,proposed a resolution adaptive quantization parameter tuning mode under the same coding complexity degree.Experiments verify that the model can effectively improve video coding quality.(3)We proposed a quantitative parameters optimization algorithm,which based on competition decision and rate distortion optimization.This paper use iterative method to improve the traditional competitive decision algorithm and to acquire MBTree algorithm performance optimal model.The algorithm is fully inherited the original MBTree algorithm that based on the idea of perception technology advantages,it effectively solves the defects of the original algorithm model,and explores MBTree algorithm can achieve maximum performance.Experimental results show that the improved algorithm,in offline case,obtained optimal model compared with the original model can significantly improve video encoding subjective SSIM,objective PSNR and MSE distortion fluctuations of frame-level and other performance.