Maximum likelihood motion compensation for distributed video coding

Stimulated by the evolution of wireless visual sensor networks and the mass commercialization of portable, low-cost video recording devices, this work presents a novel hash-based transform domain distributed video codec, offering low- complexity robust encoding. A simple hash, composed of the coarsely quantized pixel values of the downsampled frames, is employed to perform overlapped block motion estimation at the decoder, which produces a set of temporal predictors on a pixel basis. For every pixel position, a maximum likelihood estimator, based on an online estimate of the actual correlation channel statistics, combines the cluster of temporal predictors into a single value to serve as decoder-generated side informa- tion. Efficient low-density parity-check accumulate channel codes refine the side information in the DCT domain, yielding the reconstructed frames. Experimental results demonstrate that the proposed system exhibits state-of-the-art distributed coding performance. The presented video codec maintains similar encoder complexity as the state-of-the-art distributed DISCOVER codec but offers considerably superior coding performance, in particular for sequences organized in large groups of pictures or containing highly irregular motion.

Jaar van publicatie:2012

Trefwoorden:distributed video coding, Wyner-Ziv video coding, Side information generation, Maximum likelihood estimation, Hash-based motion estimation, video coding

CSS-citation score:1