Master Thesis MSTR-2011-10

BibliographyAnto Joys, Yesuadimai Michael: Effcient Context Modelling and Segmentation for Parallel Lossless Image Compression.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Master Thesis No. 10 (2011).
77 pages, english.

High speed camera systems produce large volumes of image data every second that are stored on its onboard storage. To operate these cameras over a long period of time, the data from its storage has to be transmitted to an external storage in real-time. This necessitates a fast and memory-efficient lossless image compression algorithm implemented in hardware. Although JPEG-LS, a lossless image compression scheme is considered to provide adequate compression performance at reasonable computational complexity, it makes intensive use of a context modelling scheme which typically results in an expensive (regarding speed and memory) module in a hardware implementation. The sequential processing of pixels in the raster scan order makes the speed of operation inadequate for high input data rate, e.g. 256-by-256 pixels/frame at very high FPS (frames per second). This thesis proposes a heuristic approach to improve the speed and memory efficiency of the hardware implementation by modifying the context update of JPEG-LS. Additionally to improve the speed of processing an image, a series of encoder units work on different parts of the image in parallel. The proposed modifications are tested by both software simulations in C++ and VHDL synthesis. While simulation results show compression ratio closely comparable to the original JPEG-LS, synthesis results show an improvement over JPEG-LS in terms of hardware usage and speed of operation.

Department(s)University of Stuttgart, Institute of Parallel and Distributed Systems, Parallel Systems
Superviser(s)Simon, Prof. Sven; Wang, Zhe
Entry dateMarch 25, 2020
   Publ. Computer Science