Abstract: The segmentation algorithm of triangular mesh in reverse engineering is studied. A geometric image based segmentation method is proposed. The generation method of geometric image is improved on the basis of parametrization of mesh. Combined with morphological method, two segmentation methods of normal image are proposed. One is based on wavelet reconstruction in which the approximation term is removed, and the other is based on gradient image. The threshold segmentation method of curvature image is also studied. The segmentation result of geometric image is mapped back to original mesh. Example illustrates that the geometric image based segmentation method can achieve satisfactory result for mesh data without complex free form surface.

Key words: Geometry image, Gradient image, Segmentation, Triangular mesh, Wavelets