Steps towards computer-assisted classification of colonoscopy video frames

We propose new procedures to automatically select the relevant frames of a video colonoscopy recording, saving time and computational resources and facilitating the computer-assisted analysis of the relevant aspects which appear in risk regions or regions presenting different colon pathologies. To retain only the relevant colonoscopy frames decreasing the computing time and complexity, we make use of the image properties in LAB color space, test the entropy levels of each frame and collect several other indicators useful in frame classification. In the end, a discussion on the main problems encountered and the directions towards colonoscopy automatic analysis point to the fine details needed to be solved.