Foveated Video Compression and Visual Communications Over Wireless and Wireline Networks

Friday, January 21st, 10:00 AM, ENS 602

slee@ece.utexas.edu

Abstract

In this presentation, we present a framework for foveated video compression and visual communications. We develop the following compression and communication algorithms : foveated video/image quality criterion, optimal rate control, foveation filtering, motion estimation, motion compensation, low delay foveated video transmission, foveation based error resilience over wireless networks and traffic smoothing over wireline networks. We develop a unique foveal visual quality metric, the foveal signal-to-noise ratio (FSNR). For foveated video compression, we maximize the FSNR using a new optimal rate control algorithm where a Lagrange multiplier method is defined on a curvilinear coordinate system. For the motion estimation, we define a new criterion called foveal mean absolute distortion (FMAD) and develop an efficient hierarchical block matching algorithm (HBMA). For foveated visual communications, we introduce an end-to-end system as one of future human interactive multimedia applications. For visual communications over highly error-prone mobile networks, we introduce a foveation-based error resilience scheme and an unequal error protection technique.

A list of Telecommunications and Signal Processing Seminars is available at from the ECE department Web pages under "Seminars". The Web address for the Telecommunications and Signal Processing Seminars is http://anchovy.ece.utexas.edu/seminars