Segmentation of regions in JPEG compressed medical images

A novel algorithm for the segmentation of medical images using features derived directly from JPEG compressed domain is proposed in this paper. The algorithm uses features extracted from DCT coefficients without its inverse transform and the rule based fisher discriminant K-means (FDK) technique for clustering image pixels based on derived feature vectors. In this study, we extract features for each 2×2 DCT block of compressed image. The extracted feature vector is used by an extended version of the adaptive K-means clustering algorithm for the classification of image pixels.