Research On Perceptual 3D Video Coding Technology

At present,3D(Three Dimensional)film for us is not fresh,3D TV,3D projectors and other equipment are increasingly entering ordinary families,but compared with the type and quantity of 2D videos,stereo videos in a variety of 3D devices are still few because of the huge amount of data of 3D videos,thus effective data compression is a huge challenge for researchers.The existing 3D video coding standard can eliminate the spatial redundancy and temporal redundancy of the video,but less visual redundancy.Therefore,based on the stereo video coding standard 3D-HEVC,this thesis carried out the following three aspects of research for 3D video coding technology from the perspective of human visual perception.(1)In order to reduce the encoding rate and ensure the subjective quality of the reconstructed video at the same time,from the model of visual attention and visual sensitivity information of the region of interest and combined with the way of perception threshold,this article puts forward a reflection of characteristics of human perception encoding method.Based on the region of interest extracted by the cellular automaton in the original video,and adjust the parameters of QP encoding tree unit by the Just Noticeable Distortion Encoding(JNCD)quantization threshold,in order to maintain high perceived quality of reconstruction videos and save rate.The experimental results show that compared with the HEVC reference software,the rate of the proposed algorithm is decreased by 6.93% and the FSIM_Y is improved by 0.013.(2)In order to solve the problem of insufficient binocular vision redundancy in3D-HEVC stereo video rate control algorithm,a novel 3D-HEVC rate control algorithm based on binocular masking is proposed.According to the viewpoint quantization parameter threshold model in determining the quantization parameter encoding unit level,quantization parameters of left view LCU can be obtained according to the relationship between the target bits and lambda,and quantization parameters of the right view on the corresponding encoding unit can be obtained according to the quantization parameter threshold model,the extra bit of the right view added to the left view next,until the end of encoding.The experimental results show that the average error rate of the proposed algorithm is less than 3.05%,and the rate distortion performance of this algorithm issignificantly improved compared with the HTM RC algorithm.(3)3D-HEVC rate control algorithm does not take into account the distortion of different sensitivity to this problem of the human eye with different depth in stereo video scene,this chapter proposes a control algorithm of 3D video rate disparity perception based on JND model.Firstly,the disparity perception is added to the establishment of the stereo JND model,and a stereo JND model is constructed,which combines the disparity perception and the time domain foveal feature JND model is used to adjust the rate allocation,so that the rate allocationofstereovideo is more in line with human perception.Theexperimentalresultsshowthat,comparedwith the3D-HEVCrate control algorithm,therate controlerror of this method is reduced by0.02%,and the rate distortion performance is consistent with thealgorithm.