Correntropy based sperm detection: a novel spatiotemporal processing for analyzing videos of human semen

Abstract

In this article a new spatiotemporal processing is introduced for detecting sperms in captured videos from semen specimen. In the proposed method, the temporal processing indicates primary sperms by estimating correntropy of the history sequences which are constructed for image pixels. This procedure is completed by a spatial processing that stabilizes the correct sperms by considering their connectivity. Performance of the proposed method is evaluated on several sequences captured from semen specimens containing various numbers of sperms with different shapes and contrasts. The obtained results show greater ability of the proposed method in distinguishing sperms from background of semen compared to some state of art algorithms in such way that its detection rate is at least 4% higher than the best among its alternatives without increasing false detections.

http://ift.tt/2kaKepF