logo

Regularity of B-Splines 📂Fourier Analysis

Regularity of B-Splines

Theorem1

For m=2,3,m=2,3,\dots, the B-spline NmN_{m} has the following properties.

(a) NmCm2(R)N_{m}\in C^{m-2}(\mathbb{R})

(b) For kZk\in \mathbb{Z}, in each interval [k,k+1][k,k+1], NmN_{m} is at most a polynomial of degree m1m-1.

Explicit formula of B-spline

Nm(x)=1(m1)!j=0m(1)j(mj)(xj)+m1,xR N_{m}(x) = \frac{1}{(m-1)!}\sum \limits_{j=0}^{m} \left( -1 \right)^{j}\binom{m}{j}\left( x-j \right)_{+}^{m-1},\quad x\in \mathbb{R}

Where

f(x)+:=max(0,f(x))&f(x)+n:=(f(x)+)n f(x)_{+}:=\max \left( 0,f(x) \right) \quad \& \quad f(x)_{+}^{n}:=\left( f(x)_{+} \right)^{n}

Lemma

For m=2,3,m=2,3,\cdots, x+m1x_{+}^{m-1} is differentiable up to m2m-2 times, and the m2m-2th derivative is continuous.

Proof

For m=2m=2,

x+1=max(0,x)={0if x0xif x0 x_{+}^{1}=\max(0,x)=\begin{cases} 0 & \text{if}\ x\le0 \\ x & \text{if}\ x\ge0 \end{cases}

so it is continuous at all points, and differentiable except at x=0x=0. For m=3m=3,

x+2=(max(0,x))2={0if x0x2if x0 x_{+}^{2}=\left( \max(0,x) \right)^{2}=\begin{cases} 0 & \text{if}\ x\le0 \\ x^{2} & \text{if}\ x\ge0 \end{cases}

so it is differentiable at all points. The derivative is 2x+12x_{+}^{1}, so it is continuous at all points, and differentiable except at x=0x=0.

Proof

(a)

By the explicit formula of the B-spline, NmN_{m} is a linear combination of translations of x+m1x_{+}^{m-1}. Therefore, according to the lemma below, NmN_{m} is differentiable up to m2m-2 times, and each of its derivatives is continuous.

(b)

For each j=0,1,,mj=0,1,\dots,m, the following formula holds.

(xj)+m1=(max(0,xj))m1={0if xj(xj)m1if x>j (x-j)_{+}^{m-1}=\left( \max \left( 0,x-j \right) \right)^{m-1}=\begin{cases} 0 &\text{if}\ x\le j \\ (x-j)^{m-1}& \text{if}\ x>j \end{cases}

Since NmN_{m} is a linear combination of such functions, it trivially holds.


  1. Ole Christensen, Functions, Spaces, and Expansions: Mathematical Tools in Physics and Engineering (2010), p208 ↩︎