VALIDITY-GUIDED (RE)CLUSTERING WITH APPLICATIONS TO IMAGE SEGMENTATION

When clustering algorithms are applied to image segmentation, the goal is to solve a classification problem. However, these algorithms do no t directly optimize classification quality. As a result, they are susc eptible to two problems: P1) the criterion they optimize may not be a good estimator of ''true'' classification quality, and P2) they often admit many (suboptimal) solutions. This paper introduces an algorithm that uses cluster validity to mitigate P1 and P2. The validity-guided (re)clustering (VGC) algorithm uses cluster-validity information to gu ide a fuzzy (re)clustering process toward better solutions. It starts with a partition generated by a soft or fuzzy clustering algorithm. Th en it iteratively alters the partition by applying (novel) split-and-m erge operations to the clusters. Partition modifications that result i n improved partition validity are retained. VGC is tested on both synt hetic and real-world data. For magnetic resonance image (MRI) segmenta tion, evaluations by radiologists show that VGC outperforms the (unsup ervised) fuzzy c-means algorithm, and VGC's performance approaches tha t of the (supervised) k-nearest-neighbors algorithm.