A Macrophages Image Segmentation Algorithm Based on Adaptive Region Merging and Watershed

摘要

This paper investigates the segmentation of the strong adhesion and aberrant macrophage images. We present a region based segmentation method involving the watershed algorithm and the adaptive region merging technique. Firstly, we perform the image pre-processing through top/bottom hat transformation combined with image smoothing and gray transformation; secondly, the distance transformation based watershed algorithm is used to get the initial segmentation resu finally, we design an adaptive region merging method to select different similarity criteria automatically according to the size of the adjacent areas, which will lead the initial segmentation regions merge into a final segmentation result. The experimental results indicate that the proposed method can separate the strong adhesion cells well and suppress the over-segmentation caused by aberrant cell shape distortion, which own lower error rates and are more accordant to the human vision segmentation. Moreover, the algorithm has the universal adaptability, which is applicable to the vast majority of macrophages images.