The accurate count of blood leukocytes is very important for the diagnosis of blood diseases， and the accuracy of blood leukocyte segmentation is an important pretreatment step to correctly determine the type of blood leukocytes. The B component and G component of blood leukocytes have obvious difference under different staining conditions and different lighting conditions. Accurate blood leukocyte segmentation is performed based on Otsu threshold processing and area threshold processing. Through the establishment of mathematical model， based on the morphological characteristics of intact leukocytes and lobulatednucleus leukocytes are classified and judged. The edge detection of lobulated leukocytes with mitosis is carried out， and the number of lobulated leukocytes is determined according to the maximum and minimum distance， so as to realize the accurate counting of blood leukocytes. Experimental results verify the feasibility and practicability of the proposed segmentation method.