| In recent years, there has been a dramatic growth in network-based video applications including on-demand video streaming, video phones, distance learning, and video conferencing. In this dissertation, two important aspects of wireless video communications are addressed: energy efficiency and video quality assessment. In mobile communications, energy consumption is a critical design issue. Developing techniques for efficiently utilizing transmission energy can prolong battery life, reduce interference between users, and increase the overall network capacity. We present a joint source coding and transmission power allocation scheme for wireless video communications. This cross-layer design can provide significant energy savings over approaches which consider source coding and power allocation separately.;The other primary focus of this dissertation is on methods for evaluating video quality. In packet-based communication systems, both compression artifacts and packet losses deteriorate the quality of the received signal. Traditional approaches have primarily focused on minimizing the expected value of the end-to-end distortion. We show that this type of approach may lead to undesirable perceptual artifacts. A new "variance-aware" resource allocation technique is proposed that is capable of reducing large variations in quality. Experimental results demonstrate that variance-aware resource allocation helps limit the propagation of perceptually annoying artifacts and is more robust to channel-mismatch. The concepts developed here can be utilized in a wide range of applications including error resilient encoding algorithms, and various source-network adaptation schemes. |
