Spline Approximation

Splines are piecewise-polynomial functions joined smoothly at points called knots; they approximate and interpolate functions accurately while avoiding the oscillations of high-degree polynomials.

Definition

A spline of degree k is a function that is a polynomial of degree at most k on each subinterval between consecutive knots and is continuous together with its derivatives up to order k-1 across the knots.

Scope

This topic covers polynomial splines and their smoothness conditions, the cubic interpolating spline and its end conditions, the B-spline basis that gives a stable and local representation, and the use of splines for interpolation, smoothing, and curve and surface design.

Core questions

• How do piecewise polynomials achieve global smoothness while keeping low degree?
• What determines the cubic interpolating spline, and what role do boundary (end) conditions play?
• Why is the B-spline basis preferred for representing and computing with splines?
• How do splines balance fidelity to data against smoothness in smoothing applications?

Key theories

Cubic interpolating spline

Among all twice-differentiable interpolants of given data, the natural cubic spline minimizes the integral of the squared second derivative, making it the smoothest interpolant in that sense and explaining its widespread use.

B-spline basis

B-splines form a basis of locally supported, nonnegative functions for the space of splines on a given knot sequence; they provide a numerically stable representation, a partition of unity, and efficient recursive evaluation and refinement.

Mechanisms

A cubic interpolating spline is found by solving a tridiagonal linear system for the second derivatives (or slopes) at the knots, enforcing continuity of value, first, and second derivatives, plus two end conditions such as natural or clamped boundaries. B-splines are computed by the Cox-de Boor recurrence, which builds higher-degree basis functions from lower-degree ones; because each B-spline is nonzero on only a few intervals, the resulting collocation and least-squares systems are banded and efficiently solvable.

Clinical relevance

Splines are ubiquitous in computer-aided geometric design and computer graphics (where NURBS, built on B-splines, model curves and surfaces), in data smoothing and nonparametric regression, in trajectory and path planning, and in finite-element and isogeometric analysis, because they combine local control, smoothness, and computational efficiency.

History

The mathematical theory of splines was founded by Isaac Schoenberg in the 1940s; the development of the stable B-spline representation and its recursive evaluation by Cox and de Boor in the early 1970s made splines a practical computational tool and laid the groundwork for their dominant role in geometric modelling.

Key figures

• Isaac Schoenberg
• Carl de Boor
• Maurice Cox

Seminal works

• deboor2001
• powell1981

Frequently asked questions

Why use splines instead of a single high-degree polynomial?

A single high-degree polynomial can oscillate badly between data points, whereas splines keep each piece low-degree and join them smoothly, giving accurate, well-behaved approximations even with many data points.

What is the advantage of the B-spline basis?

B-splines are locally supported, so changing one coefficient affects the curve only nearby, and they are numerically stable and form a partition of unity. This local control and stability make them ideal for design and for solving spline systems efficiently.