A Novel Auto-Segmentation Scheme for Colored Leukocyte Images

An auto-segmentation scheme for colored leukocyte images is presented in this paper. First an extracting leukocyte nuclear flow based on saturation in HSI color space is proposed. Then a number of windows containing single leukocyte are detected to count and locate leukocytes, based on a simple connected component labeling strategy. To separate overlapping leukocytes and extract single and complete leukocyte, a part of the scheme is composed of three parts: fixed threshold method (FTM), distance transformation (DT), and watershed algorithm (WA). And cytoplasm is segmented accurately by subtracting the segmented leukocyte nucleus from the extracted complete leukocyte in the same window. Experimental results indicate that the proposed scheme is effective and stable.