A novel simplification method of point cloud with directed Hausdorff distance

Three dimensional (3D) point clouds are typically used in computer vision and pattern recognition areas. In general, the raw point cloud has large numbers of redundant points which require excessively large storage space and lots of time for post-processing. This paper presents a synthetic point cloud simplification method to obtain computationally manageable point sets. First, a coarse-to-fine feature extraction manner is designed with normal vectors deviation and k-means clustering methods, which can concentrate more sample points in regions of high curvature. Additionally, the directed Hausdorff distance is performed directly on the point cloud which samples the point cloud judiciously with an edge-preserving manner. Experimental results demonstrate that the proposed method is effective for point cloud simplification, and it exhibited superior performance compared to existing techniques.