Fusion of Normalized Color and Rough Set Theoretic Approximations for Robust Color Image Segmentation

摘要

A robust technique for the segmentation of color images of natural scenes has been presented. The algorithm focuses on the problem of segmentation of images depicting regions with similar but slightly varying colors into homogenous regions, irrespective of scene geometry, by considering normalized color. For this purpose, a concept of formation of an encrustation on the histogram of normalized color coordinates has been developed. The concept is based on Rough Set Theory. The technique presents scalable levels of information details that can be utilized for the analysis. The technique has been verified by applying it to various natural images.