[Polar coordinates representation based leukocyte segmentation of microscopic cell images].

We propose an algorithm for segmentation of the overlapped leukocyte in the microscopic cell image. The histogram of the saturation channel in the cell image is smoothed to obtain the meaningful global valley point by the fingerprint smoothing method, and then the nucleus can be segmented. A circular region, containing the entire regions of the leukocyte, is marked off according to the equivalent sectional radius of the nucleus. Then, the edge of the overlapped leukocyte is represented by polar coordinates. The overlapped region by the change of the polar angle of the edge pixels is determined, and the closed edge of the leukocyte integrating the gradient information of the overlapped region is reconstructed. Finally, the leukocyte is exactly extracted. The experimental results show that our method has good performance in terms of recall ratio, precision ratio and pixel error ratio.