Image retrieval using color, texture and wavelet transform moments

Content-based image retrieval systems allow the user to interactively search image databases looking for those images which are similar to a specified query image. Similarity between images is then assessed by computing similarity between feature vectors. These features are represented in the vector form, and often are combined together. This paper explores novel wavelet approach to image retrieval based on a combination of color, texture and wavelet moments features. Color moments are used as color features which increase the precision of the retrieval process. Texture features are described by the mean, variance and energy of wavelet decomposition coefficients in some subbands. For describing shapes we propose a method that combines feature vector as a set of wavelet moment invariants.