Digital Image Compression Based on Fast Self-organizing Feature Map

摘要

A heuristic query-based algorithm is introduced to accelerate the finding of the best matching neuron (cluster center) with the smart choice, the initial states of the neurons and their desired states. A new image compression scheme based on this heuristic query-based KSOFM is presented, not only for its excellent compression quality, but also for its high compression speed. Comparison of compression quality with JPEG shows the proposed scheme provides improved performance by up to 2.03dB peak SNR over JPEG. Comparison of compression speed with a similar scheme using original KSOFM3 shows the proposed scheme improves the speed performance by 40%.