Research On Side Information In Distributed Video Coding System

With the high development of information technology and wireless network, multimedia has become one of the main means in information receiving, more and more mobile video terminals are applied in multimedia communication. The traditional video compression standards could not be applied because of its high complexity; Distributed Video Coding (DVC) has attracted considerable interests based on the new asymmetric framework. DVC adopts intra-frame encoding and inter-frame decoding, which transfers the computational burden from the encoder to the decoder, and it uses channel coding, which has the capacity against transmission errors.In DVC system, the accuracy of side information (SI) makes great influence on DVC rate distortion performance and compression efficiency, how to generate the side information accurately is one of the key problems in DVC.Based on the analysis of DVC system architecture and SI, this paper proposes a new DVC system, which is called pixel-domain and transform-domain combined DVC (PTC-DVC) system. This system is based on transform-domain DVC system, but the pixel domain processing modules are added both in the encoder and decoder. In the encoder, it extracts pixel-domain information of original Wyner-Ziv frame and transfer to the decoder, the remaining information is called transform-domain information with high correlation, the quantization loss can be reduced by processing in transform-domain. In the decoder, a SI updating solution is proposed, in the decoding pixel-domain information processing, the initial SI generated by frame interpolation is updated by the solution so as to make SI be more close to the real value. Then simulation analysis for SI updating is conducted, and experimental results show that the performance of the updated SI is better than that of the initial SI generated by frame interpolation in PSNR and perceptual quality, it brings about 2~3 dB gains in PSNR, and it helps improve the phenomenon such as blurring and ghosting in initial SI image.