A New Modification of Fuzzy C-Means via Particle Swarm Optimization for Noisy Image Segmentation

摘要

This paper presents a new clustering-based algorithm for noisy image segmentation. Fuzzy C-Means (FCM), empowered with a new similarity metric, acts as the clustering method. The common Euclidean distance metric in FCM has been modified with information extracted from a local neighboring window surrounding each pixel. Having different local features extracted for each pixel, Particle Swarm Optimization (PSO) is utilized to combine them in a weighting scheme while forming the proposed similarity metric. This allows each feature to contribute to the clustering performance, resulting in more accurate segmentation results in noisy images compared to other state-of-the-art methods.