Research And Application Of Content-Based Image Database Retrieval Technology

The rapid development of high-speed Internet and multimedia technology and the continuous improvement of mass storage devicesand processor performance have greatly promoted the popularizationand development of the graphic image system. Image database is being widely used in more and more fields like digital library, medical image databases, satellite image database, trademarks and architecture database, and so on. However, how to find the necessary image information in a large number of images is now a pressing problem in the image application.CBIR analyzes the image according to visual features of images such as color, texture, shape and spatial position relation and conducts the image retrieval by establishing a variety of image feature vector databases and building a query module.Based on a wide consultation of related research and technology materials home and abroad, the thesis, firstly, studies and analyzes the color, texture and features of shape visual underlying of the images, and builds characteristics databases by extractingthe underlying features vector of images. Then the thesis probes into the application of multi-feature integration technology to the image retrieval and presents the multiple feature based image retrieval method with the combination of other retrieval methods based on color region segmentation, shape and texture feature. The image retrieval is realized with the use of region position information and by extracting the regional color moment features from the images and by improving Hu moment invariant features as the feature vector. To solve the weak correlation between the underlying visual features and the high-level semantics of the image, the thesis introduces the relevant feedback technology for users' information retrieval purpose to improve the accuracy of the retrieval.The thesis presents a relevant feedback method based on the feature weighted, which establishes the dynamic readjustment of the weight with the combination of the feedback images marked by users so that theretrieval meets users' requests better. The effectiveness of the feedback method can be proved by experiments. The thesis designs an image database retrieval prototype system to validate the research results of the thesis.