Image Segmentation Using Fuzzy C-Means

This paper presents an implementation of image segmentation using the Fuzzy C-Means (FCM) clustering algorithm. The method assigns membership values to pixels, allowing uncertain boundaries and overlapping intensity regions to be represented through fuzzy clustering. The proposed workflow includes membership initialization, iterative cluster-centre updates, membership recalculation, crisp label generation, and evaluation using the Dice similarity coefficient and average segmentation accuracy (ASA). The implementation is considered on synthetic Gaussian images and BrainWeb T1-weighted brain MRI slices, with the aim of separating background, cerebrospinal fluid, gray matter, and white matter. The paper discusses the mathematical formulation of FCM, experimental setup, evaluation metrics, implementation considerations, and limitations of intensity-based clustering. It also identifies potential improvements, including spatial regularization, MRI bias-field correction, improved cluster-to-label matching, and three-dimensional segmentation. This work provides a baseline workflow for studying fuzzy clustering in image segmentation. Quantitative performance claims should be based on verified experimental outputs from the implementation.