Contour-based 3D point cloud simplification for modeling freeform surfaces

The reconstruction of accurate freeform surface models of 3D scanned objects is a common task encountered in creating virtual reality environments for anatomical reconstruction, cartography, cultural artifact modeling, digital archaeology, infrastructure renewal, and computer-aided design. Difficulties occur in reconstructing smooth surfaces from these scanned data sets because the acquired data is very large and is often infiltrated with scanning errors. For surface reconstruction, visualization, and interactive virtual modeling, it is necessary to reduce the amount of raw scanned data. Many existing data simplification techniques are complex and not directly applicable to spline-based surface models. A novel two stage contour-based data simplification algorithm is introduced in this paper and applied to facial surface reconstruction, which may be used for human modeling for computer games or model creation for virtual museums. The first stage extracts a series of equally spaced sectioned contours directly from a dense 3D data points. In the second stage, each extracted contour is redefined as a cubic B-spline curve by reduced number of control points defined by a user defined reduction ratio. A lofted surface is finally created from these reconstructed contours. The effectiveness of the synthetic surface reconstruction algorithm is demonstrated using its deviation values from its original point cloud data set. The experimental results show that the proposed algorithm generates a fairly accurate spline based facial model with only 5-20% of the actual scanned data, based upon the surface complexity. Performance can be improved by increasing the number of extracted contours, followed by a greater reduction ratio in the second data simplification stage.