We propose an image processing methodology to automatize the leukocyte count in Neubauer chambers. The methodology works in three stages: restoration, segmentation, and automatic count. Restoration is performed by taking local properties of neighborhoods into account. To improve segmentation, and since leukocyte concentration is low compared to other cell populations, we propose a filter of emphasis of darkest points, which is based on maximization of dissimilarity between an estimated free-leukocyte image and the respective image with leukocytes. To make the automatic count, a template based on the proposed filter is computed and used together with markercontroller watershed segmentation and k-means clustering. A discussion about the results is presented.