earticle

영어

Though clustering Analysis has been developed for many years with many clustering methods come into application, image clustering is still a difficult problem. One of the most fundamental problems is that there are many kinds of image representations, and the distinguish ability of each feature is different, so their cluster effects are also different. To decide cluster priority level of different images features on a specific image dataset, the distinguish ability of three typical image features are analyzed, and a cluster discriminant index is present, which called Simplified Overall Cluster Quality is composed of cluster compaction and cluster separation. Experimental results showed the feature with best distinguish ability also possessed best discriminant index. So this index can be used to decide the priority of features for clustering images or the best feature for image cluster.