Abstract: A new multiresolution Iterative Closest Points (ICP) algorithm is proposed by combining multiresolution features of wavelets analysis and original ICP. The first step is to triangulate the data sets, and wavelets decomposition is performed on the triangulation mesh. In order to obtain the correspondence between the data sets, the discrete curvatures of the mesh are calculated at the lowest level of wavelets decomposition. A pair of points will become the candidates if their curvatures are close enough. Wavelets reconstruction is carried out after iterative registration at each level. Some examples show that the multiresolution ICP algorithm is fast and insensitive to the noise of data sets.

Key words: computer aided geometric design, data registration, wavelets, multiresolution, ICP