Dense disparity estimation in a multi-view distributed video coding system

Distributed video coding (DVC) is a recent paradigm which aims at transferring part of the coding complexity from the encoder to the decoder. The performance of such a coding scheme strongly depends on the capacity to estimate correlation at the decoder and, consequently, on the side information quality. In this paper we consider a multi-view DVC framework and propose a very efficient dense disparity estimation technique for side information construction, based on a variational formulation. The simulation results show that our approach clearly outperforms the existing methods for inter-view side-information generation.