Research On Distributed Video Coding Based On JND

In recent years,Distributed Video Coding(DVC)technology has been a hot issue of video coding domain.By independent coding and joint decoding,Distributed Video Coding transfers the complexity of the encoder to the decoder,which reduces the encoding complexity and improves the efficiency of the encoder.The quality of side information(SI)directly affects the quality of the optimal reconstruction,which affects the performance of distributed video coding system.Hence,how to improve the quality of side information is significant for distributed video coding system.Because human eyes are the ultimate recipients of images,the image processing results should adapt the needs of human visual system(HVS).The human visual system is not able to detect the image distortion during a certain range.Redundant information which human eyes can not perceive can be discarded in the encoding process according to the visual characteristic.This paper proposed JND-based distributed video coding system,and the contents are as follows:(1)Distributed video coding system based on Just-Noticeable Distortion(JND)model is proposed.In this system,JND model is used for pre-treating the image information after transformation encoding and removing the redundant information,which human eyes are not able to detect.Only the useful information can be retained and encoded.(2)Multi-hypothesis Distributed video coding system based on JND model is proposed and Different joint conditional probability density function(pdf)are used.In this system,the quality of side information can be controlled by changing the scale factor,and the system performance with different quality side information is compared.Experimental results show that the better the information quality is,the better system performance can be obtained.Moreover,the system performance of Bayesian pdf is much better than that of weighted pdf when the quality of side information is guaranteed.