Content-based image retrieval (CBIR) technology is a hotspot of current research. The technology mainly retrieval by extract color, texture and shape visual features of images.For the shortcomings of the current retrieval method based on a single feature. This dissertation proposes a image retrieval method based on multi-feature fusion:first in order to integrate spatial information characteristics ,divide the image ,then extract color and texture feature,last IRM algorithm is used as a regional similarity matching method.For the existence of differences between System to image the bottom of the extraction of visual features and understanding of the image semantic by users.This paper presents a method which can adaptive adjust the weight of characteristics and regionals.Finally, Apply to the image retrieval method based on multi-feature fusion in the field of earthquake image retrieval in geologic prospecting field, and a CBIR Seismic image retrieval experimental system was designed. Experiments show that multi-feature fusion based image retrieval method not only enhances the flexibility of image retrieval, but also can greatly improve the efficiency of image retrieval. |