A perceptually optimized and error-resilient video codec based on 3-D SPIHT algorithm

Video transmission over unreliable networks such as Internet or wireless networks suffers from various adverse conditions such as bandwidth fluctuation, of burst-error contamination, packet loss, and packet delay due to network congestion. By taking advantage of multiple logical channels provided by digital networks, a multiple description video coding scheme is proposed to accommodate varying network environments and to meet quality of service requirements. To avoid the error propagation that usually occurred in predictive video codecs as decoding erroneous or incomplete signals, a 3-D SPIHT based coder that generates embedded bit streams is adopted. To let each description bit stream convey the most visual information with the smallest possible amount of coding bits, the wavelet coder is optimized by a color visual model that can evaluate the perceptual significance of each wavelet coefficient within subband of each color component. The error resilience in spatio-temporal domain is achieved by spatially and temporally adjacent data from other descriptions through interpolation prediction. The simulation results show that the visual quality can be maintained in transmitting video sequences at low bit rate (200 Kbps) over the channel of high data loss rate (50%).