Abstract
Capsule Endoscopy (CE) is considered an established tool for the exploration and investigation of the small intestine. There are a large number of different capsules which have been launched in the medical field by different vendors such as Given Imaging, Olympus, IntroMedic, and CapsoVision. To find experts of GI that are able to designate three to four hours for viewing one patient video will be very hard and unfeasible economically. In this research, different feature extraction techniques, such as Color Moment RGB, Color Moment HSV, Color Histogram, LBP, and Statistical features, were explored and investigated as a preprocessing phase for CE image bleeding classification algorithms. Two methods are developed using the adaptive feature selection techniques for image sequence reduction and summarization. The preliminary results showed a high reduction rate for CE images sequence size by more than 75%. Different levels of frame frequency occurrence using different features extraction techniques were developed.