146 COMPUTER AIDED ENGINEERING DESIGN
⇒
Nt
tt
tt
tt
Nt
tt
tt
tt
Nt
tt
ki
i ik
ik
i ik
ki
i ik
i
i ik
ki
i ik
,
–1, –1
–1 ––
–1,
–+1
()
=
-
()
()
⇒ Nt
tt
ttNt
tt
ki tt Nt
ik
i ik ki
i
, i ik ki
- –1 – –1, –1 –+1 –1,
() =
() +
- ()
The recursion relation to compute normalized B-splines is then
N1,i(t) = δi such that δi = 1 for t∈ [ti− 1 ,ti)
= 0, elsewhere
Nt
tt
tt
Nt
tt
tt
ki ik Nt
i ik
ki
i
i ik
, ki
- –1 –
–1, –1
–+1
() = –1,
- () +
- () +
() (5.28)
Normalized B-splines may be used as basis functions to generate B-spline curves as Hermite and
Bernstein polynomials are used in designing Ferguson and Bézier curves, respectively (Chapter 4).
In that regard, a study of the properties of B-spline basis functions becomes essential. It may be
mentioned that in some publications the notation Ni,k(t) is used instead of Nk,i(t), where k is the
degree of the polynomial in t and ti is the first knot value.
5.7 Properties of Normalized B-Spline Basis Functions
(A)Nk, i(t)is a degree k−−−−−1 polynomial in t
From Eqs. (5.20) and (5.24), Mk,i(t) is a piecewise polynomial of degreek–1 in the knot span
[ti–k,... ti) and therefore from Eq. (5.27), Nk,i(t) is a polynomial of degreek–1.
(B) Non-negativity: For all i,k and t,Nk, i(t) is non-negative
The property can be deduced by induction. In a given knot span ti–k < ti–k+1 <... < ti,
N1, i(t) = 1 for t∈ [ti–1,ti)
N1, i(t) = 0 elsewhere from Eq. (5.28)
Thus,N1,i(t)≥ 0 in [ti–k,ti)
Similarly, N1,i–1(t) = 1 for t∈ [ti–2,ti–1) and N1,i–1(t) = 0 elsewhere
Also,N1,i–2(t) = 1 for t∈ [ti–3,ti–2) and N1,i–2(t) = 0 elsewhere
Thus, both N1,i–1(t)≥ 0 and N1,i–2(t)≥ 0 in [ti–k, ti) (5.29)
From Eqs. (5.28) and (5.29),
Nt
tt
ttNt
tt
i ttNt
i
iii
i
2, i i i
–2
–1 – 2 1, –1 –1 1,
() =
() +
- () for t∈ [ti–2,ti) and N2,i(t) = 0 elsewhere.
Now, Nt
tt
i tt
i
ii
2,
–2
–1 – 2
() =
- 0 ≥ for t∈ [ti–2,ti–1]
=
0
–1
tt
tt
i
i i
≥ for t∈ [ti–1,ti)
= 0 elsewhere