Parallel program design for JPEG compression encoding

Abstract

Image compression is a kind of data compression technology. The aim of image compression is to reduce redundant information in image data. However, most image compression algorithms have problems such as computational complexity, computational load and so on. Parallel computing is an effective means to improve the processing speed. With the development of high-performance parallel processing systems, parallel image processing algorithms provides more space for improving image processing speed. And with the improvement of GPU performance, GPU is increasingly applied in the computing-concentrated data operation. According to the parallelism and programmability of CUDA, the acceleration for JPEG compression is addressed in this paper. CUDA makes it possible for GPU to do the general purpose computing. The powerful parallel computing power of CUDA GPU can improve the processing speed of JPEG image compression easily. For the parallel processing features and programmability of CUDA, this paper introduces a method of accelerating image compression based on CUDA. An optimal algorithm is proposed as well. The introduction of CUDA allows the image compression for nearly 20 to 24 times speedup. In the end of the paper, we optimize and test the program, and make the analysis of experimental results. Finally we summarize some hardware and software features of CUDA, and propose a basic method of optimizing CUDA kernels from the analysis of the experiment.