Registration of Complex Free-Form Objects from 3D Image Edge Using the Hausdorff Distance

are imposed. What characterizes our approach is that it The Hausdorff distance [6] [7] is a max-min disdoes not require explicit feature extraction, correspondtance for comparing two sets. In fact, the Hausdorff disence determination, or Surface normal estimation, an tance can be considered as a metric to measure operation that it is often sensitive to noise. (quantify) the similarity between two sets of points, i.e. the extent to which each point belonging to a model lies 2. Object registration strategy near some point of an image set. In this paper, we describe an efficient registration method for computing the rigid body transformation of complex free-form moving objects from their 3D edge images. The registration problem is formulated as an optimization problem. Our method performs the minimization of a cost function based on the Hausdorff distance. What characterizes our approach is that it does not require explicit feature extraction, correspondence determination, or surface normal estimation, an operation that it is often sensitive to noise. Therefore, the choice of the Hausdorff distance allows us to deal with outliers, occlusion, appearance and disappearance. The proposed algorithm has been tested on real data, and the results show that it is efficient, robust and yields a good transformation estimate.