A new image coding scheme based upon self organizing maps

Abstract

This paper presents a image coding scheme based upon improved self-organizing neural network, FSOM-VQ-DWT. Firstly original image is predicted by vector quantization (VQ), then the predicted error image is encoded by standard JPEG2000. To improve the performance of VQ codebook, we have proposed a new frequency sensitive self-organizing feature maps (FSOM) algorithm. Experimental results show that the proposed FSOM-VQ-DWT coding scheme can get better coding performance than standard JPEG 2000.