Research On HEVC Coding Technology Based On Visual Perception

With the rapid development of video information processing technology,the High Definition(HD)and Ultra High Definition(UHD)video have gradually appeared in our life and work.The HEVC is a newest generation video coding standard,under the condition of the same video image quality,it can reduce about 50% code stream compared with the H.264,which is especially suitable for the compression and the transmission of high resolution video.However,the high-resolution video still has a large amount of data after compressed.People pay more and more attention to the practical application effect of the video encoder in limited bit resources,and hope that the compressed video image quality can be more in line with human visual characteristics.Although the HEVC improves the overall compression efficiency,it does not encode according to the human visual characteristics.Therefore,this paper has important theoretical significance and good application value for the study on HEVC visual perceptual coding technology.Based on the characteristics of human visual system,this paper studied the HEVC coding method with visual perception feature for different video scenarios.The main tasks and the achievements are as follows:(1)The coding platform based on the HEVC coding validation model X265 is built,the coding information of the HEVC is analyzed,the overall scheme of visual perception coding is designed based on the human visual characteristics,and the evaluation criteria of video image quality is selected.(2)For the video scenes with static background,a method for extracting motion perception map based on the HEVC compression domain is proposed.Firstly,after the current frame encoded,the encoding information of motion vector,partition structure and encoding mode is extracted.Meanwhile,the quickly determine whether there is a global motion in the current frame,and calculate the amplitude of the motion vector in the current frame to obtain a motion vector magnitude map.Then,the motion vector amplitudes are processed respectively in the spatial domain and the temporal domain,therefore the motion area of the current frame can be quickly and effectively extracted.Finally,a motion perception map is generated by the motion intensityand texture intensity of the motion region,which provides a reference for the visual perception coding of subsequent motion regions.(3)For the video scenes with global motion,a texture sensing image extraction method based on edge detection is designed.Firstly,the Canny operator is used to detect the edge to obtain the binarized edge map of the current frame.In addition,t the edge points in the edge map are counted in the region of the current CU.Then,the structured texture region is processed by using the similarity of the adjacent regions in the image.Finally,the texture perception map including the current frame flat region,the structured texture region and the complex texture region are obtained,which provides references for the visual perception coding of the subsequent texture region.(4)To solve the problem of bit resource allocation in visual perception coding,a method combines the coding block quantization parameter adjustment and the frequency coefficient suppression is proposed.For the scene with stationary background,we can use the motion perception map to quantify the parameter adjustment and the frequency coefficient suppression to enhance the image quality in motion area.For the scene with global motion,the frequency coefficient is suppressed by using the texture perception maps,and the edge blocks are processed according to the direction of the motion vector.Therefore,the peak code rate and the total number of stream bits can be reduced while the video quality is guaranteed.