Medical Image Segmentation Based on the Watersheds and Regions Merging

Abstract

The accurate medical image segmentation is the basis of 3D visualization and diagnosis. A medical CT image segmentation algorithm is proposed based on watershed segmentation and regions merging to extract the liver area. The similarity criteria for the regions merging is calculated by automatically analyzing the grayscale distribution of nearby areas of the seed points selected by the user. Then, the Gaussian filter is used to smooth the CT image. And the gradient of the smoothed image is calculated by the multi-scale morphological gradient, which is the input image for Watershed segmentation. The result of the Watershed segmentation is a labeled image, and the labeled regions are merged based on the similarity criteria. Finally the region of liver is extracted by selecting the max region, and the holes in the liver area are filled, which are the vessel areas of the liver. Experimental results show that the algorithm can accurately extract the liver region in the image with little user involvement.