Segmentation of skin lesions from dermoscopic images is crucial for the early diagnosis and prognosis of many skin disorders. Fuzzy C-Means (FCM) is a widely employed partitional clustering method that has been effectively applied for image segmentation. However, FCM exhibits many limitations, including longer computational time, estimation of cluster numbers, bad convergence, local optima trapping issues, and susceptibility to noise. Therefore, researchers developed some improved variants of FCM in the image segmentation field to overcome the said issues. This study employed seven popular improved FCM variants in the skin lesion image segmentation field and performed a rigorous comparative study among them. The segmentation results of the tested FCM variants have been evaluated visually and in terms of ground truth image-based quality metrics.
Bera et al. (Thu,) studied this question.