Polar Angle Detection and Image Combination Based Leukocyte Segmentation for Overlapping Cell Images

Abstract

Leukocyte segmentation is one of the essential steps in an automatic leukocyte recognition system. Due to the complexity of the overlapping cell images, methods for leukocyte segmentation are still needed. In this paper, we first construct a combined image by saturation and green channels to extract the nucleus and in turn locate a cursory circular region of the leukocyte. Then the boundary of the leukocyte is represented by the polar coordinate. We determine the overlapping area by polar angle detection. Finally, another combined image is built based on the red and blue channels of the sub image covering the overlap to segment the leukocyte. The paper reports a promising segmentation for 60 microscopic cell images.