Retrieval of artifact images using multidimensional multiresolution features

Abstract

This paper describes a novel method of image retrieval based on multidimensional shape and blob features. Multidimensional features enable different aspects of local and global features to be reliably characterized. In addition, feature analysis is applied to multiresolution images to extract feature details at different perceptual scales. Each feature set forms a 4-D feature vector in a multidimensional multiresolution feature space. Image decomposition with the wavelet transform is used for the extraction of blobs and for multiresolution feature analysis. We apply our method of image content analysis to the classification and retrieval of images of artifacts (namely, plates, bowls and vases), and show that multidimensional multiresolution features can adequately retrieve images with a high degree of perceptual similarity.