Extracting feature points for scattered points based on gauss curvature extreme point

A fast method was proposed to extract the feature points from scattered point sets. A local surface was constructed from a spatial point and its nearest neighbors, from which the maximal point of Gaussian curvature was computed by means of gradient searching. Taking this point as start point, the maximal curvature point was searched near the area of this point. An advantage of the scheme is the local surface was fitted, at the same time, the maximal point of Gaussian curvature was computed. When the maximal curvature point was calculated, only the area of maximal Gaussian curvature was searched. It could avoid the drawback of the computing the curvature for the whole points and the time cost of comparing the maximal curvatue point. So the new scheme has higher searching efficiency.