Accurate correlation noise modeling for pixel domain distributed multiview video coding

In recent years, distributed multiview video coding (DMVC) have been proposed for the low-complexity encoding and high-complexity decoding multiview applications. The performance of DMVC highly depends on the model of the correlation noise between original frame and side information. The parameters of correlation noise model (CNM) are argued to be estimated by temporal and view correlated frames, because the fused side information is generally constructed by temporal and view key frames. On the basis of this fact, this paper proposes a practical CNM building method at decoder for pixel domain DMVC. The proposed method, according to fusion method, combines the temporal and view correlated frames to model the noise more accurately. Experiment results show that the proposed method is valid and efficiently improves the coding performance of DMVC.