Modeling correlation noise statistics at decoder for multi-view distributed video coding

Recently, multi-view distributed video coding (MDVC) receives more and more attention, as its low-complexity encoder and high-complexity decoder coding paradigm suits for many applications such as sensor networks, in which several view sequences are required to be coded by a few power-constraint encoders. Modeling the correlation noises between original frame and side information frame is a hot research issue in distributed video coding (DVC), since it is a vital factor affecting the coding efficiency. This paper firstly proposes a novel method to model the correlation noises in MDVC. And an algorithm to online estimate the model at decoder using the knowledge of adjacent views is also presented. Experiment results show that the proposed correlation model can significantly improve coding efficiency.