Search On Relevance Feedback Based Clustering And Vector Transferring

With the rapid development of information technology and the widely application of digital images, more and more complicated and large image information database come forth. How to find the image data we need fleetly and precisely is attracted by more and more people. When traditional image retrieval methods are no more appropriated for practical requirement, relevance feedback technology has been introduced and become one of the hottest research areas in image domain.The thesis studies the color-based image retrieval technique and proposes the method of image retrieval based on partition and binary on base of analysis of place information of color. The method could divide image into foreground and background precisely, give prominence to foreground in image retrieval, accord with human visual perception. Experimental results prove that the method could retrieve images more purposefully and improve precision greatly.On the basis of studying vector-transferring relevance feedback, the thesis combines vector-transferring and clustering, and increases the speed and precision of vector-transfer. Image feature vectors transfer to clustering center which they belong to more fleetly and precisely by good use of users'relevance feedback information. The arithmetic has less cost, less times of feedback and higher efficiency of image retrieval.A color-based image retrieval system which implements the technique of relevance feedback is constructed. The thesis's arithmetic are compared with each other on performance, and by the experimental data they have better retrieval efficiency.