Research Of Image Clustering Algorithm Based On Artificial Immune

With the development of multimedia technology and internet in 1990's, our life,study and work has become convenience and abundance by using multimedia information. So how to organize,express,storage,manage,query and retrieve these abundant data has been a important challenge to the traditional database technology. And then the retrieval ways of the traditional image database have been improved from test based to content based, now the image retrieval technique has gained many international successes. At the sane time, it has become a key for users' retrieval that is it reasonable to organize and manage image database, so it is very important to cluster all images in database before users' retrieving. With the improvement of images' retrieval system, it will be a farther research field that clustering based on image's emotion semantic.On the other hand, in the environment of dynamic competing many biology organisms show a great and complex ability of studying and solving some questions, which inspires human's thought, and accelerate human science technology, lately, the arisen artificial immune system has been a new application filed. Thereinto, the clustering algorithm based on immune theory can be used to cluster data containing numerical value and symbol attribute, and this algorithm has greater adaptability.Image features should be extracted and then quantified before being clustered, the data quantified will be the input data of clustering analysis.Based on above several questions, some works will be progressed in this paper. Firstly, a new feature extraction method synthesized image color and space information will be advanced, which based on introducing image's visualization features and some traditional methods of image color feature extraction, this new method can be named space histogram measurement. The main idea of this method is that first the whole color information of image will be counted using color histogram; second the space distributing information of color will be counted using centrality matrix of partitioned color; finally, the two feature vectors will be added together with some weight. emotion semantic information of some main colors will be also summarized in this paper, which is convenient for aftertime more researches' expanding.Secondly, after analyzing the limitations of traditional clustering algorithms, the clustering algorithm based on artificial immune has been proposed in this paper, and the designing steps of this new algorithm has been described in detail.Finally, four modules is contained in our experiment, namely, image obtaining,image feature extracting,immune cluster analyzing and results analyzing. In this paper, images' color feature is extracted primarily, after quantifying, these data will be the input data of clustering, at the same time, the k-mean algorithm has been used for image clustering, that the new clustering algorithm based artificial immune is better than k-mean clustering algorithm has been proved by comparing their performances, and the efficiency of image retrieval system has been improved by images' clustering before retrieving.