A Novel Methodology for Extracting Colon’s Lumen from Colonoscopic Images

Shunren Xia, Shankar M. Krishnan, Marta P. Tjoa, Peter M.Y. Goh

Recently, computer assisted diagnosis on colonoscopic images is getting more and more attention by many researchers in the world, while the colon’s lumen is the most important feature during the process. In this paper, a novel methodology for extracting colon’s lumen from colonoscopic image is presented. At first, in order to eliminate the background at the outside of colonoscopic images, an effective and easy method, which is similar to the Hough transform is used to detect the preliminary region of interest (pROI). Then the original image is segmented through two steps: relaxation process and tightening process. The relaxation process is realized by finding the all valleys from the histogram of a defined homogeneity function to produce as many homogenous regions as possible, while tightening process is subsequently employed to merge the unnecessary regions according to the color difference between them in CIE (L* a* b*) color space. After a series of postprocessing procedure, the lumen is successfully extracted. An extensive set of endoscopic images is tested to demonstrate the effectiveness of the proposed approach.

Full Text