Optimization of a NURBS representation (Q1322938)

The authors propose to compute a local optimum of the choice of control points and knot vectors for the approximation of curves and surfaces by nonuniform rational \(B\)-splines by using a penalty function that includes minimizing the distance of the control points from the geometric object and maintaining positivity of weights and monotonicity of the knot points. Their preferred method for solving the minimum problem is a Polak-Ribière conjugate gradient algorithm. They discuss various numerical aspects of computation of optimal approximation of curves and note that the same procedures for surfaces still require an unrealistically large amount of computation.