Medical image segmentation using the combination of watershed and FCM clustering algorithms

In this study, a new image segmentation technique that combines watershed algorithm and fuzzy clustering algorithms is proposed to minimize undesirable oversegmentation. Watershed algorithm invariably produces over-segmentation due to noise or local irregularities in the gradient images. In the proposed scheme, first, it presents a region merging method based on employing the Markov Random Field (MRF) model on the Region Adjacency Graph (RAG) to refine the quality of watershed algorithm, and then, the relationship of inter-region similarities is then performed by involving the spatial domain (watershed) and feature spaces (clustering) into image mapping in order to determine optimal region merging. To obtain the spatial domain and feature spaces representation of the image, spatial graph representation is used, which is derived from the watershed partitioning and feature spaces representation acquired from the Fuzzy C-Means (FCM) clustering technique. Experimental results show that the proposed technique gives more promising segmentation results in comparison with the conventional watershed algorithm by means of the assessment of several brain phantom and real data.