Hierarchical Fuzzy Min-Max Clustering Algorithm And Its Application On Image Clustering

One of the fundamental bases for content-based image retrieval is how to select appropriate multidimensional indexing algorithms to index the image feature vectors. In this thesis, image clustering technique is used to deal with this problem, and a novel Hierarchical Fuzzy Min-Max Clustering Algorithm is proposed based on Fuzzy Min-Max Clustering Neural Network for clustering image. The main work of this thesis is summarized as follows:(1) Reorganized and summarized the research findings of clustering techniques in domestic and foreign academia. And described the details of the basic concept and algorithms of clustering.(2) Studied the basic principles of Fuzzy Min-Max Clustering Neural Network (FMMCNN). Based on the analysis of shortcomings of FMMCN, a novel Hierarchical Fuzzy Min-Max Clustering Algorithm (HFMM) is presented, and the feasibility and validity of HFMM is analyzed as well.(3) Presented an overview of content-based image retrieval, and pointed that the common algorithms for image clustering bounded the number of clusters. To solve this problem, Hierarchical Fuzzy Min-Max Clustering Algorithm is applied to cluster images in this paper. And the experimental results compared with other typical algorithms demonstrated that HFMM has great ability in image clustering.