Indexing and images retrieval by content

The indexing and images retrieval by content is a very interesting topic, and recently required in our days. In this paper, we propose a new structure of indexing and images retrieval process based on feature extraction by wavelet networks moralization and intelligent method of measuring similarity. First, each query image is modeled by a wavelet network of hybrid and optimal architecture, then, for the determination of visual characteristics called low levels such as: shape, texture, color, a detection filter based on wavelet shape is proposed and a new algorithm for measuring similarity between two color images is created in which the idea is inspired by a simple set of drawing colored balls. Thus, the proposed filter is used in determining the shape descriptor and accurately in the phase of binarization of the image request, whereas our algorithm of colors is applied in the act of measuring degree of similarity of two color images. Finally, for the determination of descriptors to those closest to the query image, we propose a new technique of comparison and decision-making of membership.