A simple set of numerical invariants for the analysis of images

A new invariant, which captures some of the perceptual and conceptual features in an image is introduced. The effectiveness of the invariant is demonstrated empirically through applications to images. It is shown that extraneous objects in a texture environment can be detected and parts of a single image with distinct structural features are distinguished and separated. The invariants are not sensitive to local perturbations. Changes in the lighting or the orientation do not reduce their effectiveness. The method is based on the variation of local statistical properties of a single image and makes no use of memory or training. © 2007 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 16, 240–248, 2006