Research and System Implementation of Image Retrieval Method Based on Fuzzy Clustering
|School||Liaoning University of Science and Technology|
|Course||Computer Software and Theory|
|Keywords||content-based image retrieval fuzzy c-means clustering feature extraction similarity measurement gray level co-occurrence matrix|
With the rapid development of the Internet and multimedia technology, information on the Internet has evolved from singular text mode to multimedia modes such as images, videos, audios and flashes. The widespread application of digital images has turned into one of the major recourses in information society. It has become an urgent issue how users can obtain the most interesting images from the substantial multimedia information. Hence the development of Content-Based Image Retrieval.First, technology and current situation of CBIR are illustrated, including the key techniques such as system structure of CBIR, feature retrieval methods, approaches and standards of similarity measurement etc. Theories of fuzzy mathematics and fuzzy c-means clustering algorithm are also elaborated.Second, fuzzy c-means clustering algorithm is applied in image retrieval. To optimize retrieval speed and effect, an image retrieval method based on improved color feature and fuzzy c-means clustering is proposed. Firstly, the improved color characteristic information of each image is extracted; then, fuzzy c-means clustering algorithm is used to cluster color feature vector, and each cluster center of image class is obtained; last, the similarity between the sample image and the corresponding categories is calculated, returning the retrieval results according to the size of the similarity. Experiments indicate the proposed method has superb retrieval function with less feature dimension, shorter clustering time and quicker retrieval speed. But it doesn’t function well in images whose color is relatively plain but texture prominent. Thus, an improved algorithm is proposed that fuzzy c-means clustering algorithm is used to cluster and retrieve combining gray level co-occurrence matrix texture feature and improved color feature extracted. Experiments have proved the improved retrieval effect.Last, CBIR technology is summarized, further research fields are directed.