Unequal Loss Protection of Embedded Streams Combining Particle Swarm Optimization and Local Search

FEC-based multiple description coding (MDC) is an effective method for the transmission of scalable video stream over noisy network. However, given the transmission bandwith, it´s time-consuming to find the optimal packet number and packet length as well as reasonably allocate bits between source coding and channel coding across the packets. In this paper, we state the optimal unequal error protection(UEP) as a two-level optimization problem and use a rate-optimal solution in initial step of searching. Then, we propose a new scheme, which combines the particle swarm optimization (PSO) with the local search(LS) algorithms, so that optimal solution can be found in more efficient way. Experimental results for the 3D set partitioning in hierarchical trees (SPIHT) coders show that the new algorithm provides close to optimal average peak signal-to-noise ratio performance and that its´ running time is obviously lower than that of previously known algorithms.