Reconstruction of underlying curves with styling radius corners

This paper presents a new curve fitting framework for styling design data. Given a data set that represents a filleted-like curve, underlying curves (U-curves) and styling radius corners (SR-corners) are generated by fitting to low curvature parts and highly curved ones, respectively. A set of U-curves are firstly reconstructed as a unique C 0 composite B-spline curve, and then an SR-corner is reconstructed for each C 0 corner. This approach guarantees that U-curves can be smoothly connected through convex SR-corners while enabling full editing of the smooth corners up to sharp ones. Compared with existing schemes that naively fit a curve to each part, the proposed framework provides a guiding principle for the generation of curves, which is more suitable for styling design. Experimental results demonstrate that high-quality curves can be generated even from real-world scanned data.