An improved version of a high accuracy surface modeling method (Q381119)

The fundamental theorem of surface theory asserts existence and uniqueness (up to isometry) of a parametric surface in \(\mathbb{R}^3\) to given coefficients of first and second fundamental form, provided the first fundamental form is positive definite and the coefficients satisfy the Gauss-Codazzi equations. The surface can be recovered by solving the three Gauss equations for given initial conditions. Numerical solutions to this system of partial differential equations for function graphs are of relevance in geomathematics. Usually, the computational burdens are high and issues with numerics and memory usage have to be addressed. One possibility is to omit the mixed second partial derivative from the Gauss equations. This is, however, not satisfactory from a theoretical point of view and, supposedly, is responsible for a loss in accuracy. In this article, the authors propose an optimized difference scheme that allows to include the third Gauss equation and, at the same time, avoids critical numerics that have been observed in earlier attempts to do so. A comparison with existing methods at hand of a synthetic and a real-world example indicates better accuracy at the cost of a reasonable increase in computation time.