Research on correlation noise modeling of a generalized Gamma distribution in distributed coding

Abstract

Due to the inaccuracy of modeling the correlated noise (CN) with Laplacian distribution,the rate-distortion performance of DVC is relatively low.CN modeling via generalized Gamma (GGA)distribution was proposed for pix-el-domain DVC.Statistical features of the CN modeling were analyzed.It was found that Laplacian distribution peak was lower than actual CN distribution.GGA distribution was exploited to fit CN and the GGA distribution parameters were estimated online.The simulation results indicate that the proposed model can more accurately reflect the statistics fea-tures of CN,improve rate distortion performance,and reduce computation complexity.

Keywords

• distributed video coding;correlation noise model;generalized Gamma distribution;Laplacian distribution