An Image Retrieval Method Based On Color Histogram And Local Color Difference

With the computer image of digital technology and the rapid development ofInternet technology, now people contact more and more image information. How tosearch the needed image from the vast image database quickly, accurately, hasbecome a widespread concern topic of research, and become a key technology in theprotection of intellectual property rights, new generation of online search, digitallibrary, medical and remote sensing image processing and analysis, crime and safetyprevention and other major research projects.Content Based Image Retrieval (CBIR) appears in1990’s. At first, in the imageretrieval system, most researchers focus on signal image feature to do the retrieval job,and only use the color feature, texture feature, and figure features and so on, butalong with the development of the science and the dependent theory, the imageretrieval system begin to use the invariant theory and feedback technology, and getsome success. In this paper, we discuss the color feature and texture feature and theinvariant feature, and also explore the multi-feature fusion to do the image retrieval.In this paper, we discuss some key technology in image retrieval, and introducetwo methods of feature extraction. Explain how the two features fusioned intotogether, the experiment prove the method of image retrieval proposed in this paper iscorrect. After the theory of experiment, the paper develop the real image retrievalsystem. More and more researchers find the drawbacks of the image retrieval systemusing the signal image feature, so they begin to consider to use more features to dothe image retrieval job. And right now, the problem of how to fuse the multi-featuresand the weight of these features when to fuse into together is still difficult to solve,and do not have the standard to do these. In this paper, we do a lot of experiments,and the result of experiment shows the correction of the algorithm proposed by us,and prove the efficiency of the image retrieval use the proposed method has improvedevidently.