Automated diagnosis for segmentation of colonoscopic images using chromatic features

Development of a computer-aided scheme will assist the physician by detecting abnormalities of the colon by characterizing features of the segmented colonoscopic images. A new algorithm for segmenting the colonoscopic images based on chromatic features for computer-assisted diagnosis has been developed. The novel method exploits the local homogeneity index (/spl lambda/) definition of the pixels in a given neighborhood of the image. It is realized in two stages. Initially, segmentation is performed using /spl lambda//sub I/ in the intensity domain. A novel modified peak-finding algorithm is employed to segment the image from the corresponding histogram. In the second stage, the regions obtained in the first stage are divided into sub-regions based on /spl lambda//sub H/ in the hue domain. ne over-segmentation problem is alleviated by merging the region using color difference measures. The preliminary results support the feasibility of the proposed approach. The advantages include the analysis of both local and global information by using the /spl lambda/ histogram.