A multi-invariant difference measure for grey-scale textures

Texture is an important component in the perception, classification, identification and segmentation of images. We describe a Fourier based scalar distance measure for describing the "closeness" of the texture of images stored in a database to the texture of a query image. The measurement of distances between "normalised blurred FFT" as description of texture has proved a powerful tool in searching for perceptually similar texture within a database. The introduction of a wide range of invariants to the distance measurements has resulted in a robust comparison of texture characteristics. A satisfying result of this approach has been the perceptual similarity of the "incorrect" texture samples ranked high in the search.