Ulcer detection and diagnosis using linear coding for wireless capsule endoscopy

Ulcer is most familiar and common indication of many severe diseases in the human digestive tract. Existing techniques are not able to detect ulcer in small bowel of digestive tract. Therefore wireless capsule endoscopy (WCE) technique is widely accepted and used for diagnosis and clinical management of ulcer detected especially in small bowel. Proposed system is based on two-staged fully automated computer-aided detection system which can recognize ulcer from WCE images. Firstly, saliency reorganization method is utilized to detect ulcer candidates by segmenting image into multi-level superpixel segmentations. Then we estimate the corresponding saliency in accordance with the color and texture features in superpixel region of each level. Finally generated texture and color based saliency maps are fused together.