Motivated by the Markov chain Monte Carlo (MCMC) relaxation method of Jalali and Weissman, we propose a lossy data compression algorithm for continuous amplitude analog sources that relies on a finite reproduction alphabet that grows with the input length. Our algorithm asymptotically achieves the optimum rate distortion (RD) function universally for stationary ergodic continuous amplitude sources. However, the large alphabet slows down the convergence to the RD function, and is thus an impediment in practice. We thus propose an MCMC-based algorithm that uses a (smaller) adaptive reproduction alphabet. In addition to computational advantages, the reduced alphabet accelerates convergence to the RD function, and is thus more suitable in practice.
Anticipated extensions of this algorithm could be added to existing image and video compression standards. Taking into account that by some accounts video downloads account for over 70% of global internet data usage, fundamental improvements in lossy compression could make a significant impact. The speaker will outline ideas for future research along these lines toward the end of the talk.
Dror Baron received the B.Sc. and M.Sc. degrees from the Technion-Israel Institute of Technology, Haifa, Israel, in 1997 and 1999, and the Ph.D. degree from the University of Illinois at Urbana-Champaign in 2003, all in electrical engineering. After several years performing research in academic and industry settings, he joined the Department of Electrical and Computer Engineering at North Carolina State University in 2010 as an assistant professor, and is currently pursuing research in information theory and signal processing.