Low contrast sperm detection and tracking by Watershed algorithm and Particle filter

Sperm movement characteristic plays an important role in male fertility. Computer-aided Sperm Analysis (CASA) systems have tried to provide more accurate information about sperm motility and concentration. At the beginning of this goal, achieving correct sperm trajectory is necessary. On the other hand, sperm tracking is a challenging matter because, sperms have same size and shape, move fast also there is an uncertainty in their motions. Therefore, most of designers have attempted to reduce tracking mistakes especially, when sperms collide with each other. In This paper we try to present a sperm tracking method based on Watershed segmentation and Particle filter estimation which focuses on the sperm collision problem. In contrast of most sperm tracking algorithms based on sperm head orientation, this method tries to track sperms independent of sperm head orientation. Moreover, sperm videos were taken by common camera in order to ease availability in every laboratory. Consequently image sequences have low contrast so, processing section bears more burdens in proportion to the imaging. In this paper, the method is evaluated by tracking 10 sperms which had at least one collision along their paths. The results showed the method is robust in the cluttered scenes.