The present invention uses the content-based image retrieval preprocessing of the feature extraction process, the color (color) information and the shape (shape) information at the same time extract the color and shape characteristics of the color image using a consideration characterized vectors for the system's way and these features vectors It relates to an image retrieval method for performing matching between images. In color feature extraction method of the color images it is achieved through non-uniform quantization, and NICO (Neighborhood Information of Color Histogram) matrix and GICO (Global Information of Color Histogram) matrix creation of color images, the shape feature extraction method of the color image is image improved constant moment to consider only the entire outline of the object (improved Invariants moments: IMI) and lead (dominant) and attained a constant moment of each group of color through the calculation, an image and a query image are stored by using the extraction method in the database compared to the image search method of matching is compared to match the coefficients of each group unchanged moment of enhanced invariant moments and lead (dominant) color of the color information over the component value comparison NICO matrix and GICO matrix to match the shape information, the It is determined by the degree of similarity between the query image to each of the comparison value. Accordingly, the invention is able to achieve enhanced search efficiency even in the case of deformation of its own color histogram (histogram) caused by the change of the addition, the gamma correction value of the noise in the image.
展开▼
