Research Of Multi-feature-based Image Retrieval

This paper discusses the conception and methods of CBIR .Some future research trends are proposed also. A method for image retrieval using color feature is proposed firstly. In the method, only a small number of features, called chromaticity moments, need to be stored to effectively characterize the color content of an image.Second, aiming at the characteristics of chromaticity moments method which are considering the color distributing in the whole image merely but not considering the color distributing in the local area of the image and considering the distributing of color set in the image merely but not considering the shape, size and position of the image object, two improved methods are brought forward in this paper: one is an image retrieval method based on color-space information, it designs rectangular overlapped sub-regions to get chromaticity moments in each region. The other is a method using combined color and texture feature. The proposed pattern (local binary pattern) can depict the variation pattern of pixel gray scales in a domain to descript images texture feature. Experimental results show that combined method is efficient for image retrieval.