Universal Lossless Source Coding Techniques for Images and Short Data Sequences

Abstract: In this thesis various topics in universal lossless source coding are discussed and analyzed. The main focus in this work is on lossless data compression of grayscale still images. Such images are, for example, frequently occurring in medical imaging. Based on theoretical considerations and empirical observations new compression algorithms are presented that are, in terms of compression performance, efficient compared to traditional methods. This work includes research on how to use the Context Tree Weighting algorithm, linear prediction and probability assignment techniques in lossless data compression. The performance of these algorithms/methods is studied both asymptotically and for usage on short data sequences. The presented techniques can be used separately or together when designing efficient lossless data compression systems.