Content-based Image Retrieval

In order to make full use of images resource and get the images from the images data base we are most interested in more effectively, CBIR(Content Based Image Retrieval) has been brought up. What is called CBIR could be illustrated as following.. When retrieving images, do a feature extraction according to the content of the image to be stored and add the feature vector to the corresponding feature database, the feature of the given image will be compared to the existing features within the feature database with a certain matching method. The result of this step will be a certain number of images most similar to the image which is given. As we can see, the efficiency of the retrieval is to a certain extent determined by the quality of the feature extracted from the image.As the foundation of this paper, a general introduction is made of several key technologies involved in CBIR, including feature extraction, similarity measurement, evaluation rules and so on. Then the next part of the paper pays most attention to the feature extraction based on color histogram and presents a new method of feature extraction based on the algorithm of K-means clustering and a refined color histogram. By taking into account both global and spatial information of the image, this kind of method has made full use of the color information conveyed by the image, hence is of great advantage over other kinds of feature extraction method. Experimental results provide a good performance of the proposed algorithm. At the last part of the paper, a systematic study is made of the image retrieval based on the texture feature. And then a retrieval method using Co-occurrence Matrix is proposed. The experimental results prove the efficiency of the method.CBIR having a wide-range knowledge. In this paper, we has discuss two of the key techniques in detail,Synthesis multi- characteristics inquiry technology will be done in the future.