Surface reconstruction by a Gauss kernel integration approach

In this paper, three-dimensional surface reconstruction from the surface normal or gradient vector is studied. A nonlinear integration approach combining a Gauss kernel function based on the analysis of the numerical integration algorithm is presented. The proposed method makes use of the surface normal variation information of the points in the Gauss neighborhood and can reduce the noise significantly. The experiment results on the synthetic and real data indicate that it provides fast and reliable surface reconstruction from the gradient vectors.