Sperm morphology analysis is an essential part of the spermiogram tests. Currently, many laboratories employ visual assessment techniques whose success depend on observer experience. In this respect, many computerized approaches have been carried out in the field to eliminate the human factor in the analysis. In the proposed study, the performance of two clustering techniques have been evaluated on the segmentation of sperms in the stained images. Most of the clustering techniques are sensitive to the variations in the histogram representation. Therefore, to reduce these unwanted sensitivities, we employed a group-sparse signal denoising method named as modified overlapping group shrinkage (MOGS) as a preprocessing step. As a result of MOGS denoising, the visibility of the sperm morphology is enhanced and cleaner representations of possible blobs (sperm or non-sperm) have been obtained. After the extraction of region of interests (ROIs), sperm/non-sperm classification process has been performed by using spatial features. In order to validate the performance of MOGS denoising, the results of the segmentation techniques with/without MOGS are presented in terms of the true-positive, false-positive and missed samples metrics.