Endoscopy Video Frame Classification Using Edge-based Information Analysis

Endoscopic surgery has brought about massive changes in cardiac and thoracic surgery. The therapeutic capabilities of endoscopy provide beneficial diagnosis and treatment for physician. Although, the endoscopic surgery shows a marked improvement in diagnosis and treatment, the endoscopy video still has some frames that contain no useful information. We need to distinguish the informative frames from the non-informative frames in order to utilize information from both of them for further operations for developing Computer Aided Diagnosis systems. This paper proposes a method based on edge information analysis to distinguish informative frames and non-informative frames by utilizing the Isolated Pixel Ratio Value (IPRV). The accuracy of the proposed method was remarkably higher than that of the other edge-based technique. Results indicate that the proposed method is an effective method for informative and non-informative thoracic endoscopy frame classification.