Part-based probabilistic point matching

We present a probabilistic technique for matching partbased shapes. Shapes are represented by unlabeled point sets, so discontinuous boundaries and non-boundary points do not pose a problem. Occlusions and significant dissimilarities between shapes are explained by a 'background model' and hence, their impact on the overall match is limited. Using a part-based model, we can successfully match shapes which differ as a result of independent part transformations a form of variation common amongst real objects of the same class. A greedy algorithm that learns the parts sequentially can be used to estimate the number of parts and the initial parameters for the main algorithm.