Segmentation of Microscopic Images for Counting Leukocytes

Counting of leukocytes in peripheral blood is commonly used in basic clinical diagnosis. The reliable approach to count leukocytes depends on morphological assessment of cellular materials under a light microscopy. Microscopic evaluation by medical technologists is a time-consuming and stuffy job. It is of interest to apply computer-aided analysis with image processing and pattern recognition, instead of human to count leukocytes. In this paper, a novel segmentation method is developed to count leukocytes by incorporating textural information. The proposed method integrates a non-decimated (or called shift invariant) complex wavelet transform into watershed segmentation, and uses information obtained by adaptive threshold segmentation to refine the result of watershed segmentation. Experiments demonstrate that the proposed method can obtain satisfactory results in comparison with the judgments of medical technologists.