Recently, several researchers ([BM92], [Zha93], [CM92], [ML92], [CLSB92]) have proposed very interesting methods based on an iterative algorithm to rigidly register surfaces represented by a set of 3d points, when an estimate of the displacement is available. In this paper, we propose to introduce differential informations on points to extend this algorithm. First, we show how to efficiently use curvatures to superpose principal frame at possible corresponding points in order to find the needed rough estimate of the displacement. Then, we explain how to extend this algorithm to look for an affine transformation between two surfaces. We introduce differential informations in points coordinates: this allows us to match locally similar points. We show how this differential information is transformed by an affine transformation. Finally, we introduce curvatures in the best affine transformation criterion and we minimize it using extended Kalman filters. All this extensions are illustrated with experiments on various real biomedical surfaces: teeth, faces, skulls and brains.
CITATION STYLE
Feldmar, J., & Ayache, N. (1994). Rigid and affine registration of smooth surfaces using differential properties. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 801 LNCS, pp. 397–406). Springer Verlag. https://doi.org/10.1007/bfb0028371
Mendeley helps you to discover research relevant for your work.