Gradient estimation from irregularly spaced data sets

Topographic data sets are often composed of terrain samples arranged in irr egular meshes. Many gradient estimation methods require the data to be arra nged in a regular mesh; irregular meshes present a difficulty for them. Thi s paper presents a novel method of estimating surface gradients from irregu larly spaced inputs. The method is derived using directional derivatives an d is shown to be a generalization of traditional finite difference methods. Analytical tests were used to evaluate the method, which was shown to be a ccurate and robust.