Improved MB Mode Prediction in Extended Spatial Scalability with Error Resilient Coding

In this paper, macroblock (MB) mode prediction in extended spatial scalability (ESS) with error resilience is investigated. We extend the MB mode derivation method originally proposed by our lab to improve the efficiency of inter-layer prediction in ESS, to error prone transmission conditions. The performance of our method is evaluated in metrics such as PSNR gain and complexity reduction, which is represented by percentage change in number of MBs for each partition type at the decoder for various tested sequences. Simulation results show that significant complexity reduction is reportedly achieved with average 50% increase in the number of MBs with 16×16 mode in enhancement layer and average PSNR-Y gain of 1.35 dB at the decoder compared to JVT-W030 under testing conditions specified in JVT-V302.