Application of Watershed Algorithm for Segmenting Overlapping Cells in Microscopic Image
分水岭算法分割显微图像中重叠细胞

In order to realize automatic and fast analysis ofmicroscopic image, in this paper, the binarized image was first transformed to distance format. Fast grayscale reconstruction algorithm was then employed to reconstruct the transformed image. Finally Watershed algorithm was used to segment the result image. In thisway, the defectof the watershed algorithm that the watershed marker extracting often depends on prior knowledge of the image is avoid and the algorithm to be suitable for real clinic practice and the auto segmentation of conglutination cells is realized. At last the algorithm was programmed with C++ and was used to segment the real clinical fall off cell and pathology cell images. After being applied tomany different disease and different background cell image, result shows that the algorithm can pick up destination cells in a given image and automatically segment overlap cells accurately and rapidly under the condition of even illumination. For a given 768×576 image, the algorithm can finish the segmentation within 2 seconds in a AMD 1600+ CPU. And the segmentation result is certificated by director doctor. Thus it is feasible to apply the algorithm to the segmentation of clinical cell image.