Research On Rate-distortion Model For Transform Domain Wyner-Ziv Video Coding

Distributed video coding (DVC) features simple encoders but complex decoders,which lies in contrast to conventional video compression solutions such as H.26x andMPEG. This distinguishing feature is well suited for applications with restricted poweror storage space, such as wireless low-power surveillance and multimedia sensornetworks, wireless cameras, and mobile camera phones. Therefore, DVC is studyingwidely by the scholars around the world. Wyner-Ziv video coding has recently becomea research focus of distributed video coding.This paper presents a rate-distortion model based on Laplacian distribution (LDR-D)in unidirectional distributed video coding, which is applied to the quantization for DCcoefficient of Wyner-Ziv frame. Firstly, the rate-distortion model based on Laplaciandistribution is developed by analyzing the effects of quantization step on both distortionand rate. With the LDR-D model, two algorithms on the quantization for DC coefficientare proposed： quantization step based on LDR-D model, adaptive quantization stepbased on LDR-D model. More specifically, with the R-D model, the former uses thebinary search to find the quantization step of DC coefficient. And adaptive quantizationstep based on LDR-D model is presented to search the quantization step, which isadaptive to the content of video.Simulation results demonstrate that quantization with our proposed algorithms canhave a significant RD performance improvement. The average PSNR of WZ frames inthe system using quantization step based on LDR-D model is about0.3~1.0dB higherthan the one without using it. And compared with the former, adaptive quantization stepbased on LDR-D model can improve the average PSNR of WZ frames by0.1~0.2dBand can provide better subjective quality. While using our algorithms，the codingcomplexity increase slightly, only about4percent.