Lossless Source Coding using Cosets

According to information theory, optimal compression can be achieved by partitioning the data space into cosets and encoding a block of data with the index of the coset containing the block. This binning approach has gained recent attention in wireless sensor networks and multimedia communications research due to its inherently low-complexity encoding and error resiliency. Existing attempts offer at best asymptotically lossless compression only when the size of the block is very large. This talk presents some attempts of predictive lossless coding using cosets with small block size. A compression algorithm is first developed to encode the data individually for which rigorous mathematical analysis is performed as well. The concept is then extended to operate on a small block of data at a time using rich mathematical structures to facilitate efficient partitioning. Since theoretical analysis of these multidimensional coding schemes is extremely difficult, their efficacy is evaluated by comparing performance against conventional multidimensional predictive coding techniques. All the proposed coset-based codes inherently exploit real-valued predictions to achieve further coding gain due to avoiding rounding errors. Insights from them have led to modify the widely used conventional Rice-Golomb predictive coder achieving identical gain. Application of the proposed multidimensional codes in lossless video coding has realized compression efficiency close to the known upper bound at significantly lower computational complexity than the state-of-the-art scheme.
Dr. Manzur Murshed received his BScEng (Hons) degree in Computer Science and Engineering from Bangladesh University of Engineering and Technology (BUET), Dhaka, Bangladesh, in 1994 and PhD degree in Computer Science from the Australian National University, Canberra, Australia, in 1999. He is currently an Associate Professor and the Head of Gippsland School of Information Technology, Monash University, Australia, where his major research interests are in the fields of multimedia signal processing, multimedia and wireless communications, high performance computing, video surveillance, and simulation. He has published more than 100 refereed publications that are collectively cited by more than 400 publications according to Google Scholar and supervised seven PhD students to completion. He has been a member of the technical program committees of many international conferences and peer-reviewing papers for many reputed journals and internal conferences. He is the recipient of a academic excellence gold medal from BUET and a research excellence award from Monash University.