Efficient method for geometric attribute estimation

A new method is proposed to estimate the discrete geometric attributes, such as normal vector and curvature. The method contains two steps: points sampling and 4D Shepard interpolation. In the first step, we extract points from point cloud and estimate the attribute of the sampled points. During the second step, we interpolate a global 4D Shepard surface over these sampled points and compute the attribute of each point in point cloud. This method is very effective for using a new global surface model. It benefits many applications such as segmentation, feature extraction, data preprocessing and etc. Application examples are given to verify the proposed method.