Abstract: Scalar relationship between wavelet transforms of two subimages from two different images is discussed. Image similarity based on this relationship is defined and studied. Hierarchical wavelet transform of a local area (subimage) is demonstrated using the matrix form of wavelet transform. Based on the hierarchical wavelet transforms, linear similarity of two images of a particular type of medical images such as CT and MRI and a k-form wavelet similarity measure are discussed. The algorithm for computing similarity of two sub-images is analyzed. Numerical calculation shows effectiveness of the described similarity calculation, especially for medical images.

Key words: hierarchical image transform, image retrieval, image similarity, linear similarity, wavelets transform