3D wavelet based scalable video coding and its applications

Abstract

Scalable video coding is attractive due to the capability of reconstructing lower spatial resolution, or lower quality, or lower frame rate video signals from partial bit streams. It enables a simple solution in adaptation to different requirements from network, terminal and storage capabilities. In this thesis, a 3D wavelet based scalable video codec (SVC) is proposed through the integration of motion-compensated temporal filtering (MCTF) and JPEG2000. By exploring the unique features of MCTF and JPEG2000, the proposed simple codec not only achieves a comparable rate distortion (RD) performance but also has the properties of high scalability and error resilience. The main contributions of the developed codec lie in its simple architecture and the optimum bit allocation scheme based on the quality layer concept of JPEG2000. For MCTF based video codecs, a fundamental question is how to allocate bits to each temporal band (T-band) generated by MCTF so that certain degree of constant quality can be achieved. In this thesis, a novel approach for constant quality aimed bit allocation is proposed for the applications of adaptive stored video streaming. The basic idea of the proposed bit allocation is to adjust the energy gains to compensate the different contributions from different types of T-bands, and more or less equally distribute the distortions among the T-bands at the same level. Experimental results show that the proposed bit allocation can greatly reduce the PSNR fluctuation with only slight degradation in average PSNR. This thesis also investigates the potential usage for the developed SVC through designing two systems for streaming stored video over two types of networks, the best-effort Internet and QoS-enabled wireless network. For networks without QoS support, such as the present best-effort Internet, an end-to-end system is proposed. In particular, an end-to-end feedback control scheme is performed at the application layer to estimate the i ATTENTION: The Singapore Copyright Act applies to the use of this document. Nanyang Technological University Library available network bandwidth, which is then used to determine the video bit rate for the streaming session. For QoS-enabled networks, with the example of IEEE802.11e WLAN, a cross-layer design is proposed to optimize received video quality. Specially, at the application layer, the SVC stream is packetized with relative priority index (RPI) that marks the importance of video packets. At the network layer, available bandwidth is estimated and passed to the application layer to guide the truncation of video bit stream. In addition, adaptive QoS mapping is performed based on the RPI value of each packets. Both systems exploit the scalability property of SVC.