Research And Implementation Of Content-Based Image Retrieval Technology

With the development of the technology of the network, multimedia and database, the quantity of digital image is increasing continuously used extensively day by day. It is very important to organize, manage and utilize this multimedia information, which makes content-based image retrieval (CBIR) become one of the most active research focuses of implementation of multimedia.The CBIR establish the feature space by comprehending the content of an image and executing retrieval by measuring the similarity of images. The feature spaces usually used include color, texture, shape and position. With the feature extract method, the measure of similarity and the evaluation of CBIR are also advanced. However, there are two problems. One is that since different researchers use different feature spaces, or use the same feature space but different description of it, and the measure is different, it is difficult for retrieval to be universal, especially in WEB. The other is the single feature of image can just express single sense of visual, and it is not suitable for the semantic comprehension and complex image retrieval.In this paper, we describe the feature of image by usingISO/IEC TR 15938—MPEG-7(Multimedia Content Description Interface) and extract the 4 descriptor features including DCD(Domaint Color Descriptor), CLD(Color Layout Descriptor), HTD(HomogeneousTexture Descriptor) and EHD(Edge Histogram Descirptor). We establish, by CLD and HTD, the multi-feature spaces, which consider both color and texture, and adjust the weight of each feuatre spaces by applying MARS and GA for achieving the best result. The result of experiment shows that the retrieval time of CBIR system has decreased greatly by using MPEG-7, and the scalability and generality of the CBIR system have been achieved. The efficiency and performance of CBIR system improves greatly by using the multi-feature with optimized weight.