An Improved Watershed Algorithm Based on k-Medoids in Cervical Cancer Images

The watershed algorithm is widely used in the field of the image segmentation, which can overcome the difficulty of image analysis caused by cell overlap. However, the result of the image segmentation with the watershed algorithm were often over-segmentated. To solve this problem, the k-medoids clustering algorithm was introduced to simplify the gradient image, which is preprocessed from the original image. The edge information of the original image was obtained by the Canny edge detection operator, and the target region template was calculated by the optimized initial segmentation. Then, the segmentation result was obtained. The improved algorithm was evaluated by the segmentation accuracy compared with the professional segmented pathology image. The results show that the improved watershed algorithm proposed in this paper has a specific advantage in alleviating the phenomenon of over-segmentation, and the target area appears more completely.