Image Coding Subject to Constraints Electrical Engineering (communications Theory and Systems) Table of Contents

OF THE DISSERTATION Image Coding Subject to Constraints by Tamás Frajka Doctor of Philosophy in Electrical Engineering (Communications Theory and Systems) University of California, San Diego, 2003 Professor Kenneth Zeger, Chair An image in digital format is a rectangular array of nonnegative integers, representing the light or color intensity values in the image. Its compression or coding is concerned with trying to represent the image in as compact a form as possible. Modern image coding techniques must be able to deal with a wide variety of constraints, both from image sources and from physical system implementations. This thesis focuses on image coding under certain special conditions. In particular, the coding of images is subject to constraints on fast recognition, memory constraints from physical implementation, special content of images, constraints from the transmission media, special image types, and fidelity constraints on image resizing. We present a progressive coding technique for Region of Interest coding. We analyze image compression in terms of memory requirements and offer a low complexity approach to document image coding. We introduce a novel packet formation approach xix for image transmission over lossy packet networks. In stereoscopic image compression, we propose a method for improved coding of residual images and present an estimation algorithm for disparity window size selection. For the problem of image size conversion we introduce a solution that aims to identify the downsampling method used to improve the outcome of the upsampling operation.