Conversion and evaluation for two types of parametric surfaces constructed by NTP bases (Q2485405)

The authors focus on computational cost reduction when calculating and drawing Bézier curves and surfaces. The de Casteljau algorithm which is associated with the Bernstein basis is simple and stable, but it is not efficient enough -- its computational cost is quadratic. The authors' solution is based on the conversion of the Bézier representation into the representation given by \textit{A. A. Ball} [Comput.-Aided Des. 25, No.~8, 513--520 (1993; Zbl 0782.65012)] and generalised by \textit{J. Delgado} and \textit{J. M. Peña} (DP-Ball representation) [Comput. Aided Geom. Des. 20, No. 1, 1--10 (2003; Zbl 1069.65550)]. This representation is constructed by generalised normalised and totally positive Ball basis. Since the computational cost of the DP-Ball representation is linear, then the total computational cost (conversion including) is noticeably lower than in the case of a direct usage of the de Casteljau algorithm. In this article, the conversion formulae and conversion matrices between both the representations are derived. It can be expected that the DP-Ball representation will have a wide usage in computer aided geometric design systems, not only because of lower computational cost but also because of DP-Ball curves and surfaces have the same excellent geometric properties (shape preserving) such the Bézier ones.