Level-Set Based Point Cloud Simplification Algorithm and Its Application to Human Body Model

Simplification is an important task in computer graphics （CG）. In the light of the rapid development of reality modeling technology in CG and the 3D scanners wide application in this field, a level-set based simplification algorithm is presented, which can process high dense scan point cloud. The algorithm can solve the multi-contour and hole problems that are ubiquitous in the original point cloud. It has the advantage of ordering points automatically and holding result point cloud to keep layer-based property from original one. Level of details （LOD） human body models can be generated automatically via this algorithm by setting different simplification radius. The experiment results indicate that the method can hold model shape well under high simplification precision and the result of root mean square （RMS） error between auto-measure and manual-measure is less than angle-simplification method. The proposed method has its adaptive extension according to the requirement of high curvature detail display. The final simplified point cloud can be reconstructed by polygon or curve （surface）. The reconstruction of 3D digital human body model is the basis of virtual garment simulation and computer animation research. The various precision constructed human body models using the proposed method can be applied in further research.