Detection of Bleeding in Wireless Capsule Endoscopy Images Using Range Ratio Color (1005.5439v1)

Abstract: Wireless Capsule Endoscopy (WCE) is device to detect abnormalities in colon,esophagus,small intestinal and stomach, to distinguish bleeding in WCE images from non bleeding is a hard job by human reviewing and very time consuming. Consequently, automation for classifying bleeding frames not only will expedite the process but will reduce the burden on the doctors. Using the purity of the red color we can detect the Bleeding areas in WCE images. But, we could find various intensity of red color values in different parts of the small intestinal,so it is not enough to depend on the red color feature alone. We select RGB(Red,Green,Blue) because it takes raw level values and it is easy to use. In this paper we will put range ratio color for each of R,G,and B. Therefore, we divide each image into multiple pixels and apply the range ratio color condition for each pixel. Then we count the number of the pixels that achieved our condition. If the number of pixels grater than zero, then the frame is classified as a bleeding type. Otherwise, it is a non-bleeding. Our experimental results show that this method could achieve a very high accuracy in detecting bleeding images for the different parts of the small intestinal