Improved virtual channel noise model for transform domain Wyner-Ziv video coding

Distributed video coding (DVC) has been proposed as a new video coding paradigm to deal with lossy source coding using side information to exploit the statistics at the decoder to reduce computational demands at the encoder. A virtual channel noise model is utilized at the decoder to estimate the noise distribution between the side information frame and the original frame. This is one of the most important aspects influencing the coding performance of DVC. Noise models with different granularity have been proposed. In this paper, an improved noise model for transform domain Wyner-Ziv video coding is proposed, which utilizes cross-band correlation to estimate the Laplacian parameters more accurately. Experimental results show that the proposed noise model can improve the rate-distortion (RD) performance.