Abstract
In this thesis, we present two hierarchical GOP (Groups-of-Pictures) coding schemes for supporting VCR-like functions and error resilience in video streaming systems. For supporting VCR-like functions, our approach is based upon reshaping the ordinary linear encoded GOP structure into a hierarchical binary tree encoded GOP structure to reduce the transmission overhead caused by frame dependencies. The resulted video streaming system can immediately provide all directions video playback with any speed-up factor. We additionally propose a proximity-based approximation playback scheme along with a modified depth first search (MDFS) algorithm that can further reduce the bandwidth requirements for performing fast playback operations without causing visual quality degradations evidently. For supporting error resilience, we reorganize the regular linear GOP structure into a hierarchical double-binary tree. The proposed scheme can effectively recover a lost frame and prevent error propagation phenomenon. For a single frame loss, we guarantee that both of its next and previous frames still can be successfully decoded, thus we have sufficient temporal and spatial information to reconstruct the damaged frame. Our new coding structure even can recover the successive lost frames. The experimental result showed that the video quality has a graceful degradation when the loss rate increases rapidly. Comparing with the conventional GOP, PSNR values can be improved from 0.5 dB to 3 dB. A set-top box architecture that supports the hierarchical GOP structured video streams playback is also described in this thesis.