Master Thesis MSTR-2010-05

BibliographyTantawy, Haitham Assem: Parallelization of JPEG-LS Encoder for Lossless Compression of Images in CUDA.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Master Thesis No. 5 (2010).
66 pages, english.
Abstract

This thesis investigates the tendency of encoding parallelism of the lossless JPEG-LS encoder. We analyze the existing data dependencies arising from the context modeling process used in the standard, and which dictates a sequential encoding of pixels that are classified to the same context. Previous work, found in the literature, has approached parallelism in JPEG-LS either by pipelined hardware implementations or software implementations, which modify the context update procedures in order to break the data dependency loop, and thus become no more compliant to the established standard. In our work, we present an encoder design, referred to as round encoder, which achieves parallelism in JPEG-LS by encoding the image in terms of context rounds. Each round contains pixels of different contexts which can be encoded in parallel. The proposed round encoder design accomplishes the same number and order of the performed context updates like the standard, and thus is fully compatible. The degree of maximum achievable parallelism however depends on the context distribution of pixels in the image, and thus parallelism is not guaranteed. The thesis also presents a parallel software implementation of the JPEG-LS encoder, which realizes the round encoding concept being fully compatible with the standard. A GPU is used as a target hardware architecture for our implementation in conjunction with a CPU for the sequential working parts. A speedup is achieved for some of the parallelized algorithmic blocks. However, no speedup is obtained for the whole parallel JPEG-LS encoder with respect to the original sequential implementation. Possible reasons behind the low performance are discussed with further suggestions for future work.

Department(s)University of Stuttgart, Institute of Parallel and Distributed Systems, Parallel Systems
Superviser(s)Simon, Prof. Sven; Wahl, Simeon
Entry dateMarch 2, 2020
   Publ. Computer Science