Approach to clustering large visual databases using wavelet transform

Many applications demand the capability of retrieval based on image content. A classification mechanism is needed to categorize images based on feature similarity. An effective classification of the images can support efficient retrieval of images. In this paper, we investigate a feature-based approach to image clustering and retrieval. Four different texture-based feature sets of images are extracted using Haar and Daubechies wavelet transforms. Using multi- resolution property of wavelets, we extract the features at different levels. The experimental results of our clustering approach on air photo images are reported.