Matching: Invariant to translations, rotations and scale changes

We present an approach to invariant matching. In this approach, an object or a pattern is invariantly represented by an object-centered description called an attributed relational structure (ARS) embedding invariant properties and relations between the primitives of the pattern such as line segments and points. Noise eeect is taken into account such that a scene can consist of noisy sub-parts of a model. The matching is then to nd the optimal mapping between the ARSs of the scene and the model. A gain functional is formulated to measure the goodness of t and is to be maximized by using the relaxation labeling method. Experiments are shown to illustrate the matching algorithm and to demonstrate that the approach is truly invariant to arbitrary translations, rotations, and scale changes under noise.