Research On Visual Feature Based Image Retrieval Technology

Along with the arrival of multimedia time, people can get more and more image information. How to provide an efficient approach to query these images which have rich content therefore becomes an active hot point in the area of retrieval research. Content based image retrieval (CBIR) is just one of key technologies for such a problem.Abstracting image's content in the level of feature is the base of CBIR. So, the technology of image retrieval based on the visual feature is researched in this paper. Both the color image and trademark are our research objects. For the former, color is one of the most important visual features of image and compared with the shape feature, it can be get more easily. According to this, we present two retrieval methods in which color feature is used, one is based on image's dominant color and the other is based on image's shape. As for the trademark, we divide it into two kinds, ordinary trademark and combined trademark. A retrieval method using entropy and moments is proposed for the former, and another method based on both shape feature and spatial relationship is for the latter. The two different ways have the same purpose, that is, to make image's part feature and whole feature supplement each other. We also experimented with all the methods mentioned above in several image data bases and the results are satisfied.