An image classification method based on a SK sub-vector multi-hierarchy clustering algorithm

To make image databases more effectively organized, we present a SK (sequence clustering plus the K-mean clustering) sub-vector multi-hierarchy clustering algorithm in this paper. It clusters images to numbers of classes automatically, according to human perception. It utilized HSV histograms, wavelet texture features, color-texture moments, a gray gradient co-occurrence matrix, and hierarchical distribution features, to put similar semantic images into the same set. The algorithm was effectively proven by experiments for automatic image clustering, for the clustering image categories have semantically similar properties. It could organize image database efficiency so as to improve image retrieval.