Data compression for thin client computing

Thin client computing has gained wide acceptance in the past few years, and recent advances in broadband connectivity have lead to a growing interest in thin client computing across the Internet. Thin client computing is a model in which an entire desktop is executed on a server and its user interface is displayed on a lightweight client. While the simplest remote display protocol represents changes to a desktop as variable-size images, it requires sophisticated data compression to enable interactive graphical logins across the Internet.; This dissertation presents novel lossless compression methods for sequences of images that mirror a desktop on a remote display. We exploit the observation that a typical desktop contains substantial amounts of rendered text, capture pixel-level redundancy and code large blocks of pixels that already occur on the thin client's screen by a copy operator. While pattern matching and substitution is generally expensive, we specialize textual image compression for synthetic images and detect block motion by tracking relatively unique features in a probabilistic manner. Our methods compress exceptionally well because they exploit redundancy at the pixel-level as well as repetitive micro- and macro-blocks.; We develop a novel measurement methodology for evaluating the performance of bandwidth-adaptive thin client systems and thoroughly analyze the ability of our methods to transport the graphical output of real benchmark applications across the Internet. Our analysis demonstrates that our methods compress better than existing state-of-the-art methods and substantially improve the usability of thin client computing across the Internet.