Vector quantization for image compression based on fuzzy clustering

In this paper a codebook design for image compression based on the fuzzy c-means (FCM) algorithm is presented. The codebook design from training vectors is viewed as a fuzzy clustering problem of unlabeled data points into clusters. Due to computational cost of FCM to generate the codebook, a fast version (FFCM), which operates on the image histogram, is used to obtain a good initial codebook to start the FCM algorithm. Experimental results are presented to illustrate the performance of the proposed compression method.