Study On Technology Of Perceptual Video Coding And Error Concealment

In recent years,the perceptual video coding technology based on human visual characteristics has been paid more attention in the field of video compression.It has become an important research field in computer science and human visual engineering.This paper attempts to use the human visual attention model and human visual sensitivity model to improve the compression efficiency of the video.The transmission error may occur when the compressed video data transfers in the unstable channel,the video decoding error will affect the subjective effect.Therefore,the error concealment technique to improve the robustness of video bitstream has become a hot research direction in video coding technology.In this paper,The temporal and spatial correlation of video is used to repair the video data in order to improve the subjective and objective quality of the video.Main works include in the paper:Firstly,this study proposes a just noticeable distortion(JND)model based on saliency map detection algorithm that combines visual sensitivity model and visual attention model to use for video coding.The proposed model combines the advantages of the pixel domain and transform domain of JND model,which is proposed for the video texture,boundary and flat region respectively.Meanwhile,a salience algorithm based on Weighted Sparse Coding Framework is used to detect the salient regions in the video to adjust the JND model,and finally to remove redundant data from the video.The experimental results shows that the average bitrate reduction is 8.82%at the same subjective quality compared with HM16.0.Secondly,this thesis explores a perceptual video signal processing model based on structure analysis.First,the order structure component,disorderly structure component and texture component can be estimated by some method,which are used to JND model.Then the disorderly structure uncertainty of disorderly structure component is available to adjust spatio-temporal saliency regions of frames and contrast masking.Finally,the spatio-temporal saliency regions are used to adjust the JND threshold.Experimental results show that the 9.43%bitrate reduction is achieved in our system compared with HM 16.0 without jeopardizing the perceptual quality.Finally,this thesis explores a error concealment algorithm based on spatio-temporal information.The algorithm contains spatial error concealment(SEC)and temporal error concealment(TEC).In the spatial domain restoration,the SEC strategy is used to select the suitable spatial method by location information and Intra Mode of surrounding received macroblocks.In the temporal domain restoration,the inter partition mode of surrounding macroblocks are used to divide the corrupted macroblock,the several temporal concealment models is used depending on the partition mode of the corrupted block.Experimental results show that the proposed algorithm has a significant improvement in both objective quality and subjective quality compared with traditional methods.