Efficient inter-layer motion compensation and error resilience for spatially scalable video coding

Due to the heterogeneous wired or wireless networks, video spatial scalability has become important in video compression and transmission. Spatial scalability is designed for video applications where customers with various device resolutions and network conditions receive different presentations of the same video content. In this thesis, efficient motion compensation techniques are proposed for spatially scalable video coding, where intra-layer temporal redundancies as well as inter-layer spatial redundancies are exploited in the enhancement layer encoding. The proposed approach selects the method with the lowest rate distortion cost from pyramid motion compensation, subband motion compensation and intra-layer independent encoding for each macroblock of the enhancement layer. Our results show that the proposed method outperforms the spatial scalability performance of the scalable H.264/AVC extension, i.e., the SVC standard. An adaptive scheme of the inter-layer motion compensation is also proposed. It automatically selects either subband motion compensation or intra-layer independent encoding at a sub-block level. Experimental results demonstrate that the adaptive approach improves the coding performance with a decrease in encoder complexity.;To understand the efficiencies of these spatial scalability techniques, we further derive the rate distortion lower bounds of the pyramid and subband motion compensation from rate distortion theory. We also analyze their performances based on quantization noise modeling. Numerical evaluations of those rate distortion functions show that the inter-layer pyramid and subband motion compensation methods are better than intra-layer inter coding if the base layer is encoded at a relatively high quality or the motion estimation accuracy is low in the enhancement layer. Experimental results from real video data encoding also show that the presented theoretical analysis can be very useful to understand the efficiencies of these spatial scalability techniques.;Besides scalability, reliable transmission of video over error prone channels is also important. We combine the spatially scalable video coding with multiple description coding techniques to make the scalable bitstream more resilient to packet erasure. Our evaluations demonstrate that the error resilience performance of the spatially scalable bitstream with multiple descriptions is significantly improved compared to the single description case.