Multi-resolution surface reconstruction

This paper presents a novel surface reconstruction algorithm, which can directly generate multi-resolution meshes from the unorganized cloud points. Local normal change is used to determine whether local surface is flat or rough. According to the user-specified normal change threshold, a min-max box containing all sample points is split into octree cells with different sizes. If the local surface is flat, a portion of the isosurface will be extracted in a large cell, otherwise in a small one. Thus the proposed algorithm can reduce the number of triangles in mesh compared with conventional Marching Cubes (MC) algorithm. The datasets of Terracotta Warrior and Horses in Qin Dynasty and the Kettle in Song Dynasty are used to verify the efficiency of our algorithm. The experimental results show that this algorithm call greatly decrease the number of triangles with the increase of the normal chan-e threshold and maintain the details of surface.