Method as a preprocessing stage for tracking sperms progressive motility

Abstract

Methods of human semen assessment are quite wide ranging. In this paper, we use background subtraction methods in order to detect progressive sperms whose quality of movement strongly influence fertility. Robust Principal Component Analysis (RPCA) is a powerful algorithm which has been used recently for background subtraction purposes. Sperm tracking problem can also be defined as a background subtraction problem. In RPCA algorithm, data is represented by a low rank plus sparse matrix. In our approach, the foreground data is recovered through such matrix decomposition. We compare the RPCA approach with four other background subtraction methods in order to check accuracy of algorithm as a preprocessing stage in sperm tracking. Two basic background subtraction methods of approximate median and frame difference have been examined. Furthermore, another more recent method of mixture of Gaussian model and robust probabilistic matrix factorization have been used for comparison. As the results show, the RPCA approach is more robust and less sensitive to outliers in comparison with other background subtraction methods.