Research On Color Feature Based Image Retrieval

With the rapid development of the computer technology, communication technology, multimedia network technology and automation technology, there are more and more resources of figure images, and the application of the image is becoming so extensive in the latest few years. How to organize and utilize the image sources that becomes more and more massive and how to search the image quickly and effectively, has become an problem to resolve in the figure image domain. All the problems have promoted the research and development of image retrieval technology. Content-based Image Retrieval (CBIR) becomes most active one in multimedia retrieval field.Because the technology of Text-based Image Retrieval (TBIR) needs a great deal of manual work and has low nicety in search matching, CBIR has emerged to be one of the hot research areas in image domain in order to satisfy with the retrieval demand of immensity image source the technology always analyses the color, texture, shape, and other low-layer image features, to establish retrieval vectors as retrieval index. CBIR system compares with the features of the key image and other images in the database, shows the result to the user. In the technology of CBIR, the problem how to extract the visual features and which kind of features would be extracted is a core. Although many feature-extracting methods and similarity measurements have been raised, they are not mature enough to get excellent retrieval results. New technology of CBIR researching is required to be improved.In this paper, on the base of widely referring to the material, including the reports, books and papers about Content-Based Image Retrieval of home and abroad, the whole evolution of the theory on the Content-Based Image Retrieval technique and development are reviewed. We introduce the traits of CBIR, TBIR and SBIR (semantic-based Image Retrieval), then evaluate them. We mostly discuss some issues in the CBIR. At the beginning, we analyze the systems in the net, and show thepresent situation and difficulty of CBIR, and make the study of the technology of CBIR, especially, the feature-extracting methods. In order to improve inquiry accuracy rate of image retrieval system, an improved algorithm based on color feature is proposed; The way of parting is used in the improved algorithm based on color feature in order to embody space information of color and improve the effect of retrieval .Based on the first method be proposed in this paper , a new feature-extraction method of image retrieval based on global and parting histogram is proposed. It assures the stability of image when it is revolved and embodies space information of color. Weight value can be adjusted by system according to characteristics of images in order to show the important of some child blocks to obtain the best effect of image retrieval.In our process of system implement, RGB is translated into HSI when we deal with the color information, for HSI is more close to human vision. In addition, we divide H into 12 shares, range is from 1 to 12; we divide S into 3 shares, range is from 0 to 2, so does I, in order to reduce the computation and improve computational efficiency. We transform three-dimensional color space vector into one-dimensional feature vector. And we adopt Euclidean distance as the corresponding similarity-measure method in our process of system implement.After taking the improved algorithm based on color feature and selecting the model of color space and similarity-measure method , we finish the designs of the experiment image retrieval system based on color feature in the VC++6.0. The system realizes the image retrieval function. Finally,we have compared the results of the three methods.The fulfillment of this paper realizes the effective extraction and retrieval of image color feature. It also provides us with theory method and practice technology. And the system of CBIR, designed in the experiment, will provide with reference value in the future of CBIR technology.