Trimming the Wireless Capsule Endoscopic Video by Removing Redundant Frames

Wireless Capsule Endoscopy (WCE) is a new challenging clinical technique for obtaining a detailed video of the small intestine. A major drawback of this technology is that it generates a large volume of video which will take time for visualization. In this paper, we propose a method to reduce the visualization time by removing those redundant video frames. There are two types of redundant frames. The first one is called non-informative frames, which is created by turbid fluids, foods, and faecal materials. The second one is called redundant frames, which is caused by the non-uniform movement of WCE. We first applied Gauss Laguerre Transform (GLT) to detect the bubble-like non-informative frames. The redundant frame was detected by a distance measure based on image edges. We apply Canny operator with a fixed smoothing parameter to detect edges of the image. An asymmetric distance was defined to measure the shift of edges from one image to another. The image whose edge shift to its neighboring image is smaller than a threshold was defined as a redundant frame. Experimental results demonstrate an average of 17% of non-informative frames and 37% of redundant frames for one person, without a loss of useful frames.