Parallelization of an image compression and decompression algorithm based on 1D wavelet transformation

Abstract

Wavelet analysis has received considerable interest in the recent years because of its efficiency in the several practical applications. Image processing for wavelet transformation is considered as one of the most powerful methods that provide a good quality of results. However, its implementation may be too time-consuming accordingly to the problem size. Parallel processing can be a solution to speed up wavelet transformation programs. In this context, and in order to have a quick image compression/decompression program based on 1D wavelet transformation, we have designed three parallel algorithms that where implemented on an IBM RS6000/SP machine. The first parallel algorithm exploits control parallelism it was developed with OpenMP and executed on one four-processor node. The two others exploit data parallelism and were developed with MPI directives. Finally, we present an evaluation of these algorithms based on an experimental study.