One of the most important outcome predictors of malignant tumors is the mitotic count, i.e. the division speed of cells. This value is computed from the patient's tissue samples by medical experts, that count each mitosis case one by one under a microscope, and as such it is a time consuming process. In order to accelerate it, we present in this paper a system capable of mitosis detection from histological breast cancer images. To this end we have developed a fully automatic solution based on mathematical morphology. The proposed approach has achieved the 10th best performance among 14 teams at the international mitosis detection contest organized by ICPR'12.