Affiliations: Department of Electronic and IT Media Engineering, Seoul National University of Science and Technology, Seoul, South Korea
Note: [] Corresponding author: Gilwon Yoon, Department of Electronics and IT Media Engineering, Seoul National University of Science and Technology, 232 Gongneung-ro, Nowon-gu, Seoul 139-743, South Korea. Tel.: +82 2 970 6419; Fax: +82 2 979 7903; E-mail: gyoon@seoultech.ac.kr.
Abstract: In this study, we propose an image analysis method for locating bleeding sites of the gastrointestinal-endoscopic image. Endoscopic images are particularly difficult to analyze due to variations in illumination, various shades of blood color, noise and presence of food debris. In this study, blood was identified using the probability distribution models based on optical characteristics of blood. Blood discrimination capability was enhanced by taking morphological features such as clustering into account. In terms of noise removal, we proposed a so-called switching adaptive median filtering method and it provided better results compared to the median filtering or low pass filtering. Each pixel in the image was analyzed and its sensitivity and specificity were computed. For 15 images with bleeding, a sensitivity of 86.71% and a specificity of 88.01% were achieved.
