Wireless Capsule endoscopy video summary

Wireless Capsule endoscopy (WCE) has been developed to allow direct view of the entire small bowel in human body in a noninvasive manner for the first time. But the visual inspection of a large number of video frames produced in each examination poses a tedious task to physicians. In this paper, we propose a novel scheme aiming for reduction of the number of frames in a video so as to partially solve this problem. To achieve this goal, the original WCE video is first divided into different segments by detecting video boundary among the video sequence. Key frames from different segments are then extracted to summarize the video with k-means clustering. Preliminary experiments on our present video data verify that it is promising to employ the proposed scheme to abstract a WCE video.