"nonlinear Model-based Analysis and Description of Images for Multimedia Applications" Deliverable 2 Invariant Gray Scale Features Technische Universitt at Hamburg-harburg

The paper presents techniques for the construction of invariant gray scale features. That are characteristics of gray scale images which remain constant if the images are transformed according to the action of a transformation group. We rst introduce a general method applicable to arbitrary transformation groups and specialize the results afterwards to the group of image rotations and translations, the a ne group and the group of projective image transformations. It is shown that each of these three groups has speci c characteristics which are decisive for the properties of the resulting features. It has to be emphasized that the invariant gray scale features developed in this paper are not geometric invariants which rely on a parametric description of the object boundaries and on the calculation of appropriate derivatives. In fact, our algorithms only require the segmentation of the objects from the background and the pixel coordinates together with the associated gray values. Some experimental results concerning the invariance properties, the robustness and the separation properties of the invariant gray scale features are presented. Chapter