Journal of Information Hiding and Multimedia Signal Processing c 2012 ISSN 2073-4212 Ubiquitous International Volume 3, Number 1, January 2012 Progressive Coding and Side Information Updating for Distributed Video Coding Mohamed Haj Taieb Laval University, Quebec, QC, Canada, G1K 7P4 mohamed.haj-taieb.1@ulaval.ca Jean-Yves Chouinard and Demin Wang Laval University, Quebec, QC, Canada, G1K 7P4 Communications Research Centre Canada, Ottawa, ON, K2H 8S2 Canada jean-yves.chouinard@gel.ulaval.ca, Demin.Wang@crc.ca Khaled Loukhaoukha and Gr` egory Huchet Laval University, Quebec, QC, Canada, G1K 7P4 Communications Research Centre Canada, Ottawa, ON, K2H 8S2 Canada khaled.loukhaoukha.1@ulaval.ca, gregory.huchet@crc.ca Received October 2010; revised July 2011 Abstract. Distributed video coding is a research field which brings together error cod- ing techniques along with video compression methods. It is usually based on a Slepian- Wolf encoder which often involves turbo codes because of their strong error correction capabilities. Typically, the turbo encoder generates parity bits which are sent to re- fine the side information reconstructed at the decoder by interpolation of the already received neighboring key frames. In this paper we introduce a novel distributed video coding scheme with progressive decoding. The side information is updated progressively as long as the current frame is being decoded. The proposed architecture considers a chess-board struc- ture for block grouping. A subset of blocks are first sent, decoded and then used to update the side information. Then, the remaining blocks are sent and de- coded using the up- dated and more accurate side information. The implementation of the progressive coding shows an improvement up to 1.7 dB over the conventional DVC architecture. Keywords: Distributed video coding, bidirectional motion estimation and compensa- tion, progressive coding, Wyner-Ziv. 1. Introduction. Distributed video coding (DVC) is a relatively recent research field which began attracting several researchers since the introduction of Stanford University DVC architecture in 2002 [1]. The improvements achieved in the subsequent studies, are mainly based on the same architecture. The frames of a video sequence are divided into intraframes (key frames) and interframes (Wyner-Ziv frames). The intraframes are sent and used by the receiver to generate side information to decode the interframes. This side information is actually the result of an interpolation, or an extrapolation, of the received key frames to estimate the WZ frames. This estimation is done without any knowledge about the current frame and makes several assumptions during the motion compensation. In this paper we present a modification to this commonly used architecture to allow for progressive video coding and decoding. A WZ frame is divided into blocks and the blocks are grouped into two sets of blocks. One of them is sent first to the receiver and decoded 1