An Adaptive Aco-based Fuzzy Clustering Algorithm for Noisy Image Segmentation

The fuzzy c-means (FCM) has been a well-known algorithm in machine learning/data mining area as a clustering algorithm. It can also be used for image segmentation, but the algorithm is not robust to noise. The possibilistic c-means (PCM) algorithm was proposed to overcome such a problem. However, the performance of PCM is too sensitive to the initialization of cluster centers, and often deteriorates due to the coincident clustering problem. To remedy these problems, we propose a new hybrid clustering algorithm that incorporates ACO (ant colony optimization)-based clustering into PCM, namely ACOPCM for noisy image segmentation. Our ACOPCM solves the coincident clustering problem by using pre-classified pixel information and provides the near optimal initialization of the number of clusters and their centroids. Quantitative and qualitative comparisons are performed on several images having different noise levels and bias-fields. Experimental results demonstrate that our proposed approach achieves higher segmentation accuracy than PCM and other hybrid fuzzy clustering approaches.