Bleeding Detection in Wireless Capsule Endoscope Based on Color Feature Vector

Abstract

Wireless capsule endoscopy (WCE) is an electronic revolutionary technique with no pain, easy to use and provides the capability of inspecting the whole small intestine which traditional ways do not. However, WCE produces so many pictures of a patient that it burdens on the work of doctors detecting diseases. To solve this problem, a highly promising automatic method based on the color feature of CIELAB and CMYK color space is proposed. The first step is image preprocessing, which extracts the region of interest from the image. Then, according to the characteristics of the CIELAB color space, color balance based on gray world assumption is executed and the images’ local contrast is improved. The second step is to generate a binary vector. The third step, we use a support vector machine (SVM) as a classifier to identify images and use RBF kernel function to check the feature vector’s performance. Experimental work shows that the proposed approach improves precision.