Journal of Information Organization Juan C. Arcila-Diaz, Victor Tuesta-Monteza, Heber I. Mejia-Cabrera,Maria P. Trujillo-Uribe, Kim Jeong-Hyun 8 1 2018 http://www.dline.info/jio/fulltext/v8n1/jiov8n1_3.pdf Colonoscopy videos contain blurred, non-informative frame sequences due to the rapid movements of the endoscope during the exploration which need to be excluded to allow the expert physician to carry out his work in less time. In this paper two methods of artificial vision are proposed for the automated extraction of informative frames based on detectable characteristics of them. The first method allows for frame sorting into informative and non-informative based on the number of contours detected in each frame. The second method makes use of the dense optical flux to determine the percentage of individual frame motion, for group the frames whit K-Means algorithm by your motion in three groups: mean motion (informative frame), large motion and little motion (non-informative frame). Both methods were successful in filtering out blurred frames from colonoscopy video samples with the first method outperforming the second, i.e., 76.7% accuracy versus 74.7%, respectively.