Effectiveness of a strip-mining approach for VQ image coding using GPGPU implementation

GPGPU (general purpose computing on graphic processing unit) attracts a great deal of attention, that is used for general-purpose computations like numerical calculations as well as graphic processing. As the Internet grows, the amount of data transmitted on the Internet increases dramatically, so the data compression has been more important than ever. The VQ (vector quantization) compression is one of the most important methods for compressing multimedia data, but the execution time of the codebook generation for the compression is very large. In this paper, we focus on the codebook generation algorithm called PNN (pairwise nearest neighbor) and consider the availability of GPGPU by using CUDA (compute unified device architecture) for the algorithm. Our experimental results show that the speedup of more than 40 is achieved with a strip-mining approach on NVIDIA's GPU compared with a conventional core 2 processor. We also evaluate it from the viewpoint of the power consumption.