Researching Image Retrieval Methods Based On Shape And Space Relationship

With the development of the network technology, all kinds of vocations use images more widely, so the image-management and retrieval are more important than before. For images, labeling is taking time and energy, and can't be labeled using one or several words, also for many images, manual labeling is impossible. In order to use the images effectively and better, the impactful retrieval methods are unavoidable, so Content-Based Image Retrieval is advanced, that is CBIR.Content-Based Image Retrieval is using image-character such as color, texture, shape and space relationship, and in virtue of image-processing technology, pattern recognizing technology, computer vision technology to research. So far, Image Retrieval based-on one character is being researched widely, but the result is not ideal. So Image Retrieval based-on two or more characters is attracted more attention.This thesis first researched the actuality, developing trend and application of Image Retrieval based-on shape and space relationship, then in light of shape and space relationship are the core content of the image and the important basis for people to identify and sort images, also Image Retrieval based-on shape and space relationship is proved feasible, so the paper put forward Image Retrieval methods based-on shape and space relationship. First to research image character of shape and space relationship, second to find one or more depicting methods, through character extracting, character depicting, character matching to retrieval. This paper used two character depicting methods: Fourier method and SVD (Single Value Decomposition) method; and used two matching methods: distance matching and SVD matching; and used recall, precision and time complexity to validate the methods. The results proved that SVD retrieval method and SVD matching method are better than Fourier method and distance matching method.This paper also brought forward another matching method——DP matching method, and using more experiments to prove it is feasible and valuable.This thesis also set up a retrieval system based on retrieval project above. The system has the advantages of observing easily, retrieving intuitionisticly,conveniently, it can hold out browsing retrieval and demonstration retrieval.