The problem of computing a piecewise linear approximation to a surface from its sample has been a focus of research in geometry modeling and graphics due to its widespread applications in computer aided design. In this paper, we give a new algorithm, to be called offset surface filtering (OSF) algorithm, which computes a piecewise-linear approximation of a smooth surface from a finite set of cloud points. The algorithm has two main stages. First, the surface normal on every point is estimated by the least squares best fitting plane method. Second, we construct a restricted Delaunay triangulation, which is a tubular neighborhood of the surface defined by two offset surfaces. The algorithm is simple and robust. We describe an implementation of it and show example outputs. The problem of computing a piecewise linear approximation to a surface from its sample has been a focus of research in geometry modeling and graphics due to its widespread applications in computer aided design. In this paper, we give a new algorithm, to be called offset surface filtering (OSF) algorithm, which computes a piecewise-linear approximation of a smooth surface from a finite set of cloud points. The algorithm has two main stages. First, the surface normal on every point is estimated by the least squares best fitting plane method . Second, we construct a restricted Delaunay triangulation, which is a tubular neighborhood of the surface defined by two offset surfaces. The algorithm is simple and robust. We describe an implementation of it and show example outputs.