Algorithms for low bit rate visual communication systems

Due to the advancements in the field of VLSI and digital communications, visual communications over low bandwidth channels has been made possible. An example of such a system is video telephony, which allows audio-visual communication through the public switched telephone networks. In this dissertation, we focus on the problem of designing low bit rate visual communication systems. Solutions to this problem must contend with two major limiting bottlenecks, which are low bandwidth and the amount of computational power. We study the various components that make up such a system and propose possible solutions that meet the design requirements. Specifically, we consider the following issues: motion estimation and compensation, edge sensitive image scaling, rate control, and video codec optimization.; We propose a scheme for motion estimation and compensation that is based on two-dimensional mesh motion estimation. The proposed method improves upon the traditional 2-D backward motion estimation by improving motion prediction for patches containing motion discontinuities.; To address the bandwidth limitation, we propose an edge-adaptive video scaling unit that will be placed at the receiver side. The insertion of the video scaling unit enables the display of larger sized image while a small sized image is transmitted over the low bandwidth channel. The goal of this approach is to reduce the source rate while minimizing the loss of perceptual quality by paying special attention to the edge regions.; We proposed a new predictive rate control method for the H.263 video encoder. The scheme has a low computational requirement and it is suitable for real time rate control. Efficient bit allocation is achieved by taking into account the activity of the macroblocks in the frame to be encoded. Simulation results show the new scheme performs more accurate bit allocation than the TMN rate control scheme without sacrificing the quality of the encoded video.; We consider the problem of software optimization of video codecs on a specific hardware platform. Specifically, we design and implement of a fast H.263 video encoder on the Pentium with MMX Technology processor. The various components of the encoder were chosen with the tradeoffs between compression efficiency and speed. We observe that the selection of fast components should not be based on the number of arithmetic operations when a specific hardware platform is considered. (Abstract shortened by UMI.)...