Exploiting spatial correlation in pixel-domain distributed image compression
David Varodayan, Anne Aaron, Bernd Girod · 2006
Abstract. Contrary to convention, we construct distributed image compression codecs that operate in the pixel-domain, yet exploit spatial correlation at the decoder only. For lossless compression of binary images of text, we propose two novel decoders: one assumes the image to be a one-dimensional stationary Markov process, and the other assumes it to be a two-dimensional stationary Markov random field. We demonstrate that these decoders enable compression approaching their respective Slepian-Wolf limits and they perform better than the baseline pixel-domain decoder by factors of at least 2 and 4, respectively. Index Terms Slepian-Wolf coding, distributed source coding, Low-Density Parity-Check codes, Baum-Welch algorithm 1