A Hybrid Image Segmentation Approach Using Watershed Transform and FCM

Abstract

Medical image segmentation is an essential step for most subsequent image analysis tasks. In this paper a hybrid image segmentation algorithm is proposed, which combines the morphological method of watershed and fuzzy c-means (FCM) clustering. A dilation-erosion contrast enhancement approach is used as a preprocessing stage in order to obtain an accurate estimation of the image borders. Then an initial partitioning of the image into primitive regions is produced by applying the maker-controlled watershed transform. After edge post-processing, the regions' statistical characters are inputted to a FCM clustering process for the final segmentation. Merging the watershed regions through the FCM clustering obtains a better initial setting from the preceding steps, accelerates convergence speed, and improves the accuracy of segmentation. The hybrid algorithm is applied to lung extraction in computerized tomography (CT) images. The experiments show that the algorithm is more effective for medical image segmentation than FCM and watershed algorithm.