Computer Aided Design

Outline


Interpolation and polynomial approximation

Interpolation

Lagrange

Cubic Splines
Approximation

Bézier curves

B-Splines

1
 Computer Aided Design


Some vocabulary (again ;)

Control point : Geometric point that serves as
support to the curve

Knot : a specific value of the parameter u
corresponding to a joint between pieces of a curve

Knot sequence : the set of knots values (in an
increasing order).

Wolfram research
2
 Computer Aided Design


Interpolation

The curve passes through the control points

Approximation

The curve doesn't necessary passes through the
control points

But these have an influence ...

Statistic approaches ?

Least squares

« Kriging »

Not very adapted to geometric modelling

3
Computer Aided Design

Interpolation

4
 Computer Aided Design
Interpolation

Interpolation

We want draw a regular and smooth parametric
curve through a certain number n of points Pi

Several families of base functions are available

Most obvious are polynomials

There are others -

Trigonometric functions (by mean of a Fourier decomposition for
instance)

Power functions

etc...

5
 Computer Aided Design
Interpolation


We must choose the parametrization (nodal
sequence)

Uniform u2=2

u1=1

u0=0
un-1=n-1

6
 Computer Aided Design
Interpolation


Or non-uniform u2=7

u1=2.5

u0=0
un-1=15

7
 Computer Aided Design
Interpolation


Can we choose u as a curvilinear abscissa ?

In principle no since we don't know the final shape
of the curve beforehand (with the exception of
interpolation points)

We will see later that it is often impossible that u
corresponds with s exactly along the whole curve
using analytical functions.

But nothing forbids to get close to that –
numerically...

8
 Computer Aided Design
Interpolation


Parametrization as an approximate arc length

u2=d1+d2

d2
u1=d1

d1 u0=0
dn-1
un-1=d1+...+dn-1

9
 Computer Aided Design
Interpolation


In all cases, we want to have :
xu i = xi
P u i =P i ≡
{
y u i = y i

We are going to interpolate the functions x(u) and
y(u) with ONE polynomial with n parameters

This one must be of order p=n-1 :
n−1
2 n−1 j
x u=a 0a 1 ua 2 u ⋯a n−1 u =∑ a j u
j=0

We set the linear system and solve...
x u 1 =x 1

{ ⋮
x u n−1 =x n−1 10
 Computer Aided Design
Interpolation


Vandermonde matrix
2 n−1
1 u0 u ⋯ u
a0 x0

( )( ) ( )
0 0
x u 1 =x 1

{ ⋮
x u n−1 =x n−1
1 u1
⋮ ⋮
u
⋮
2
⋯ u
⋮
2
1
⋮
n−1
⋮
1 u n−1 u n−1 ⋯ u n−1 a n−1
n−1
a1 = x 1
1
⋮
x n−1
Can be solved by classical numerical methods, but ...

The condition number of this matrix is VERY bad
 It must be solved for each RHS member ( ( xi ) or ( yi ) )
, or have to take the inverse of this matrix, or perform
an LU decomposition.

11
 Computer Aided Design
Interpolation


Instead of setting the polynomial in x and in y
and solving, we can put it under the following
form : n−1

{
p
x u=∑ x i l i u n−1
0
n−1
⇔ P u=∑ P i l ip u
p 0
y u=∑ y l u
i i
0

, where the lip(u) are a polynomial basis of order p=n-1.


These polynomials verify, for an interpolation :
p
l u j = ij
i

We have only one computation to do for any
position of the interpolation points, knowing the ui
(the parametrization) 12
 Computer Aided Design
Interpolation

 Lets determine the li(u) : they are of order n-1

and such that : l p u =1 ; i= j
l ip u j = ij ⇔
{ i
p
j

l u j =0 ; i≠ j
i

lip may be factored by (u-uj) for j≠i

p
l u=k u−u 0 u−u 1 ⋯u−u i−1 u−u i1 ⋯u−u n−1 
i
k is such that l ip ui =1
k =ui −u 0 ui −u 1 ⋯ui −u i−1 u i −u i1 ⋯ui −un−1 −1
n−1
p (u−u j )
l (u)=
i ∏
j=0, j≠i (ui −u j )

The lip are the Lagrange polynomials of order p

13
 Computer Aided Design
Interpolation


Lagrange polynomials
p n−1
p l u u−u j 
l u= p '
i = ∏
l u i u−ui  j=0,i≠ j u i −u j 
n−1
p
l u=∏ u−u j 
j=0
n−1
p'
l (u i )= ∏ (u i −u j ) (first derivative w/r to u at u=u i )
j=0 , j≠i


Properties
n−1
Interpolation Partition of unity l
p
∑ i (u)=1
l ip u j = ij n−1
0

x(u)=∑ x i l ip (u) with x i =1∀i

0 14
 Computer Aided Design
Lagrange polynomials


Order 4
 The ui are evenly
distributed between
u=0 and u=1

The sum is equal to 1
(partition of unity)

Presence of negative
values

15
 Computer Aided Design
Lagrange polynomials


Order 10

Presence of huge
overshoots near the
boundaries

16
 Computer Aided Design
Lagrange polynomials


The interpolation is represented by the following
form :
n−1 n−1
p p (u−u i )
P (u)=∑ P l (u) with l (u)=
i i i ∏
i=0 j=0, i≠ j (u j −u i )


Two things worth noting:

The curve depends linearly on the position of the
points

It is formed by a weighted sum of basis functions
that express the influence of each point on the
curve 17
 Computer Aided Design
Lagrange polynomials


An experiment

We approximate a circle by an increasing number
of points

Simultaneously , approximation order increases
 In every case, the curve is C

18
 Computer Aided Design
Lagrange polynomials

3 points, order 2 (a parabola !)

19
 Computer Aided Design
Lagrange polynomials

5 points, order 4

20
 Computer Aided Design
Lagrange polynomials

11 points, order 10

21
 Computer Aided Design
Lagrange polynomials

21 points, order 20

22
 Computer Aided Design
Lagrange polynomials


Does it work ?

The points were set exactly on the circle

What occurs if their position is inaccurate ?

Or if the approximated shape is not so simple ?

We are going to see two cases

The coordinates of the points are perturbed randomly

A deterministic increase and decrease of the radius

23
 Computer Aided Design
Lagrange perturbation


Random perturbation

Each point is moved radially by a value between
-0.5 and +0.5 % of the circle's radius

24
 Computer Aided Design
Lagrange polynomials

3 points, random perturbation 1%

25
 Computer Aided Design
Lagrange polynomials

5 points, random perturbation 1%

26
 Computer Aided Design
Lagrange polynomials

11 points, random perturbation 1%

27
 Computer Aided Design
Lagrange polynomials

21 points, random perturbation 1%

28
 Computer Aided Design
Lagrange polynomials


Runge phenomenon

Similar to Gibbs phenomenon of the
decompositions in harmonic functions

29
 Computer Aided Design
Lagrange polynomials


Deterministic perturbation

Each point is moved radially by a value of -5 or +5
% of the circle's radius

30
 Computer Aided Design
Lagrange polynomials

3 points, deterministic perturbation 10%

31
 Computer Aided Design
Lagrange polynomials

5 points, deterministic perturbation 10%

32
 Computer Aided Design
Lagrange polynomials

11 points, deterministic perturbation 10%

33
 Computer Aided Design
Lagrange polynomials

21 points, deterministic perturbation 10%

34
 Computer Aided Design
Lagrange polynomials


How to minimize Runge's phenomenon ?

The problem here is the use of a unique polynomial
and regular intervals between knots.

Instead, if we concentrate knots at the extremities,
the interpolation is less prone to Runge's
phenomenon.

Make use of Chebyshev knots:
2i −1
n
u =cos
i
2n( π ) in the interval [-1,1]

or uin = 1 + 1 cos 2i−1 π

( ) in the interval [0,1]
2 2 2n 35
 Computer Aided Design
Lagrange polynomials

21 points, random perturbation 1%

Uniform nodal intervals – lagrange interpolation

36
 Computer Aided Design
Lagrange polynomials

21 points, random perturbation 1%

Using Chebychev knots

37
 Computer Aided Design
Lagrange polynomials


Perturbation of a point

We shift one point by a big displacement

38
 Computer Aided Design
Lagrange polynomials

11 points, using Chebychev knots

39
 Computer Aided Design
Lagrange polynomials

11 points, using Chebychev knots

40
 Computer Aided Design
Lagrange polynomials

99 points, using Chebychev knots

41
 Computer Aided Design
Lagrange polynomials

499 points, using Chebychev knots

42
 Computer Aided Design
Lagrange polynomials


Even without perturbations, we still do
numerical approximations

Up to how many points can we pass through ?

43
 Computer Aided Design
Lagrange polynomials

51 points (lagrange polynomial of order 50)

44
 Computer Aided Design
Lagrange polynomials

61 points (lagrange polynomial of order 60)

45
 Computer Aided Design
Lagrange polynomials


Here it has nothing to do with Runge's phenomenon

It's rather an accumulation of errors due to
arithmetic calculations done in finite precision

In our case, near the boundaries, we do an inexact sum
of very big numbers in absolute value, but whose sum
(the exact sum) is close to the unit (partition of unity).
n−1 n−1
p p (u−u i )
P (u)=∑ P l (u) with l (u)=
i i i ∏
i=0 j=0, i≠ j (u j −u i )

Big numbers in absolute value,

computed with a good relative accuracy

Sum increasing the relative error

46
 Computer Aided Design
Lagrange polynomials


How to evaluate a polynomial ?

Under its Lagrange form (here) : robust except at the
boundaries

Alternative : newton's formula and divided differences

Horner's scheme
n−1
P u=∑ i u i ⇔ P u=a 0ua 1 ua 2u a n−1 
i=0

Optimal method with respect to the number of arithmetic operations
for a polynomial expressed as a sum of monomials – may be
numerically unstable

Others techniques adapted to particular forms of
polynomials later in the course e.g.

Algorithm of De Casteljau for Bernstein polynomials

Algorithm of De Boor for the evaluation of Bsplines

In any case, one should be cautious with numerical errors !47
 Computer Aided Design
Numerical errors


Numerical errors

Floating point calculations are (most of the time)
inaccurate

Analysis of the rounding done during operations in
floating point :
x⊖ y= x− y 11  , ∣ 1∣≤2 
x⊕ y= x y 1 2  , ∣ 2∣≤2
x⊗ y= xy1 3  , ∣ 3∣≤

Mathematically equivalent calculations but
expressed differently give distinct results

An example with x 2 − y 2 = x y x− y
48
 Computer Aided Design
Numerical errors


Error made with the expression  x y x− y
 x⊕ y⊗ x⊖ y  =  x− y 1 1  x y1 2 13 
=  x y  x− y 1 1 2  3 1  2  2  3 1  3 1  2 3 
≈ 5


No catastrophic increase of the relative error

Error made with the expression x 2 − y 2
( x⊗ x)⊖( y⊗ y) = [ x 2 (1+ δ1 )− y 2 (1+ δ 2 )](1+ δ3 )
= (( x 2− y 2 )(1+ δ 1)+ (δ1−δ 2 ) y 2 )(1+ δ 3 )
2 2 2 2
= (( x − y )(1+ δ1+ δ3+ (δ1−δ2 ) y + δ1 δ3 + (δ1 −δ 2 ) y δ 3 ))

When x is close to y, the error can be of the order of
magnitude of the theoretical result... 49
 Computer Aided Design
Numerical errors


Some useful rules (not a comprehensive list !)
2 2

Prefer  x y x− y to x −y

Lagrange form more accurate than
N
horner's scheme …

E. g. sum of many terms S =∑ X [ j ]
j=1

Naive algorithm :
S=0;
for (j=1;j<=N;j++){ S=S+X[j] ; }
return S ;
involves an error ≈ N 

Kahan's summation algorithm
S=X[1];C=0
for (j=2;j<=N;j++)
{ Y=X[j]-C; T=S+Y; C=(T-S)-Y; S=T }
return S ;

involves an error ≈2  50
 Computer Aided Design
Numerical errors

Example of catastrophic rounding
nx*ny samples

Computation of an integral :
ymax

S= ∫ f ( x , y)dxdy with f (x , y)=x + y

Ω
2 2


nx−1 ny−1
S≈ ∑ ∑ f  x i , y  j det J
i=0 j=0 ymin
xmin xmax
dx= x max− x min / nx y
x
dy= y max− y min / ny
x i = x min i dxdx / 2
y  j = y min j dydy /2
det J =dx dy 51
 Computer Aided Design
Numerical errors

Computations made with the following
parameters:
xmin= ymin= 0.0 ; xmax= ymax= 1.0 ; nx = ny = 10, Sexact=2/3
1) Single precision floating point numbers
2) Double precision floating point numbers
3) Quad precision floating point numbers
4) Single precision floating points numbers with
Kahan's summation algorithm
bechet@yakusa:floating_error$ ./test 10
1) sum (float )=0.66500002145767211914
2) sum (double )=0.66500000000000025757
3) sum (ldouble)=0.66499999999999999997
4) sum (kahan )=0.66499999999999999997
Note : Program should be compiled without optimization ! 52
 Computer Aided Design
Numerical errors
bechet@yakusa:floating_error$ ./test 100
1) sum (float )=0.66664981842041015625
2) sum (double )=0.66665000000000051994
3) sum (ldouble)=0.66665000000000000093
4) sum (kahan )=0.66664993762969970703

bechet@yakusa:floating_error$ ./test 1000

1) sum (float )=0.66668075323104858398
2) sum (double )=0.66666649999999805232
3) sum (ldouble)=0.66666649999999999982
4) sum (kahan )=0.66666668653488159180

53
 Computer Aided Design
Numerical errors
bechet@yakusa:floating_error$ ./test 10000
1) sum (float )=0.36836880445480346680
2) sum (double )=0.66666666499985449690
3) sum (ldouble)=0.66666666499999997623
4) sum (kahan )=0.66666656732559204102

bechet@yakusa:floating_error$ ./test 100000

1) sum (float )=0.00390625000000000000
2) sum (double )=0.66666666665538221181
3) sum (ldouble)=0.66666666665000055245
4) sum (kahan )=0.66666662693023681641

Compensated summation algorithm coming from :

William Kahan. Further remarks on reducing truncation errors. Comm. ACM, 8(1):40,1965.
54
 Computer Aided Design
Numerical errors

y=p(x)=(1-x)n
for x=1.333
and 2<n<41

S. Graillat, Ph.
Langlois, N.
Louvet
Compensated
Horner Scheme
Research Report
RR2005-04, LP2A,
University of
Perpignan, France,
july 2005
55
 Computer Aided Design
Numerical errors

Definition of the condition number of a
numerical expression

Ratio of the direct error to the inverse error

∣ y∣
K P , x=lim sup
 
  0  x∈ D  ∣ x∣ inverse
x p y= p( x )

error δ x

For a polynomial ̃p
δy
x +δ x
direct
( monomial form ) p
error

n i ̃y = ̃p ( x )= y +δ y
∑i=0 ∣ai∣∣x∣ = p( x +δ x )
K  P , x= n
∣ a
∑i=0 i x i
∣ 56
 Computer Aided Design
Numerical errors


There are various compensated algorithms to
carry out calculations on floating point numbers.

Ex. Kahan summation, Compensated Horner
scheme ...

In general, they allow to have similar results as
when using internal floating point with a
precision twice that of the input data, following
by a final rounding.

See references available on the course's
website for the compensated Horner scheme
and other documents.
57
 Computer Aided Design
Lagrange polynomials

Morality :

Lagrange interpolation is not suited beyond 10's of
control points because of the Runge phenomenon

A modification of the position of a control point
leads to global changes of the curve.

The evaluation of high order polynomials expressed
as monomials leads to numerical problems.

No control of the slopes at the boundary of the
curve (start and finish).

65
Computer Aided Design

Splines

66
 Computer Aided Design
Splines


Motivation
Let us imagine that we have many (100's of)
control points

But we don't want a Lagrange interpolation !

We should stay with a low order scheme but
conserve enough freedom to pass through every
point

Curve defined by pieces ... and of low order (1)

67
 Computer Aided Design
Splines


We are going to build a low order interpolation
for each knot interval, such that we can impose
slopes at the knots.

P'=P'i
P=Pi+1
u=ui
P'=P'i+1
P=Pi
u=ui+1

68
 Computer Aided Design
Splines


In each range [i,i+1] , we want to have an
independent polynomial

We have 4 parameters : position at each knot
and associated tangents.

The basis must have 4 degrees of freedom, thus be
of order 3 in the case of polynomials.
x (u i )=x i

{
P (ui )=P i ≡ y (u i )= y i
...

x [i ] u= A[i ]0  A[i ]1 u A[i ]2 u2  A[i ]3 u3 , u∈[ui , u i1 ]

69
 Computer Aided Design
Splines


First, every interval has a unit length i.e.
ui1 −ui =1

Then we ensure identical intervals [0...1] between
each interpolation point :
u−u i d u
u = =u−u i =1
u i1−u i du


On each interval i , we thus have the following
relation:
2 3
x[i] ( u
̄ )=a[i] 0 + a u
[i]1 ̄ +a u
[i ]2 ̄ + a u
[i] 3 ̄ ,u
̄ ∈[0,1]

70
 Computer Aided Design
Splines

P'=P'i
P=Pi+1
P'=P'i+1 u
 =0
P=Pi
u
 =1

71
 Computer Aided Design
Splines


We pass through both control points:
2 3
P ( ū=0)=P i ⇔ a [i ]0 +a [i ]1 ū+ a[i ]2 ū + a[i ]3 ū =x i
2 3
P ( ū=1)=P i+1 ⇔ a [i]0 +a [i]1 ū+a [i] 2 ū +a [i ]3 ū = x i+1

We impose both slopes :
P ' ( ū 0 =0)=P 'i ⇔ a[i ]1 +2 a [i ]2 ū+3 a [i ]3 ū 2= xi'
' 2 '
P ( ū=1)=P 'i+1 ⇔ a[i ]1 +2 a [i] 2 ū+3 a[i ]3 ū = x i +1

At the end :
a[i]0 = xi

{ a[i]1 =x 'i
a[i] 2=3 xi1− x i −2 x'i −x 'i1
a[i]3 =2 xi −x i1  xi' x 'i1 72
 Computer Aided Design
Splines


We have continuity

We have continuity of the derivatives

But how to choose the slopes ?

Let the user choose ( “artistic” freedom)

Automatically ...

73
 Computer Aided Design
Splines


By finite differences with three points :
' xi1− xi xi − xi−1
x=
i 
2u i1−u i  2u i −u i−1 

At the boundaries, we use finite differences (asymmetric)
' x1 −x 0 ' x n−1− x n−2
x=
0 x n−1 =
u 1 −u0 u n−1−u n−2

The result depends on the parametrization !

Cardinal spline x '0=(1−c)( x 1− x 0 )
' x i1− xi−1
x =1−c
i , 0≤c≤1 '
x n−1 =(1−c)( x n−1 − x n−2 )
2

c is a « tension » parameter. c=0 gives yields the so called
“Catmull-Rom” spline, c=1 a zigzagging line.
74
 Computer Aided Design
Splines

Continuity of the curve/parameter

but loss of regularity (and of geometric continuity
in many cases)

5 points, finite differences by varying the parametrization

[0..1] , [0..2] , [0..5] , [0..10]

75
 Computer Aided Design
Splines

5 points, Cardinal Spline (Catmull-Rom) c=0

76
 Computer Aided Design
Splines

Catmull-Rom Splines are widely used in computer graphics


Simple to compute, effective

Local control (price to pay : discontinuous secd derivative)

Animations with keyframing

Ensures a fluid motion because of the continuity of the slope

77
 Computer Aided Design
Splines

5 points, Cardinal Spline c=0.25

78
 Computer Aided Design
Splines

5 points, Cardinal Spline c=0.5

79
 Computer Aided Design
Splines

5 points, Cardinal Spline c=0.75

80
 Computer Aided Design
Splines

5 points, Cardinal Spline c=1.0

81
 Computer Aided Design
Splines


We can impose the continuity of second derivatives...

On a curve with n points, we have n extra relations to
impose

We may impose the continuity of the second derivative only
on the n-2 interior knots
What about the 2 points on the boundary ?

Impose a vanishing second derivative.
We obtain what is called « natural spline »

We could also impose the slopes (i.e. only n-2 relations remaining)

Or , impose that the third derivative is zero on the points 1 and n-2

That means a single polynomial expression for the first two knot
intervals, and the last two.

82
 Computer Aided Design
Splines


Natural Spline : mathematical approximation of the
spline historically used in naval construction.

C. DeBoor
83
Computer Aided Design
Splines

Boeing
84
 Computer Aided Design
Splines


We impose the continuity of the second derivatives
''
x[i−1] 1=x ''[i ] 0⇔ 2a[i−1] 26a [i−1]3 =2a [i ]2

We substitute in the “internal” equations
2[3 x i − x i−1 −2 x 'i−1 − x 'i ]6[2 x i−1− x i  x 'i−1 x 'i ]
' '
=2[3 x i1 − x i −2 x i − x i1 ]

Finally we obtain :
' ' '
x i−1 4 x  x
i i1 =3 xi1 −x i−1 

85
 Computer Aided Design
Splines


At the boundaries we want
'' ' '
x [0] 0=0⇔ 2a[0] 2=0 2 x0  x1 =3 x 1− x0 
'' ' '
x [n−2] 1=0⇔ 2a[n−2] 26a[ n−2] 3=0 x n−2 2 x n−1=3 x n−1− x n−2 


We have then a linear system with n unknowns :
'
x0

   
2 1 3 x 1− x0 

 1 4 1 x '1 3 x 2− x0 
1 4 1 x '2 = 3 x 3− x1 
⋱ ⋮ ⋮
1 4 1 x' 3 x n−1− x n−3 
n−2
1 2 ' 3 x n−1−x n−2 
x n−1
86
 Computer Aided Design
Splines


By solving the system, we have :
x '0


a [i]0 = xi

{
'
x 1 '
' , which is substituted in a [i]1 =x i
x 2 ' '
a [i] 2=3 xi1− x i −2 xi −x i1
⋮ ' '
'
x n−2 a [i]3 =2 xi −x i1  xi x i1
x 'n−1

,to get the polynomial in each portion :

x [i ]  u =a [i] 0a[i ]1 u a [i ]2 u2 a[i]3 u 3 , 0≤u 1
From the global parameter u, we have to find in which portion we are ( the
value of i ) , then compute right polynomial... 87
 Computer Aided Design
Splines
Catmull-Rom spline

5 control points, natural spline

88
 Computer Aided Design
Splines


Another experiment

We approximate a circle by a number of increasing
points

Simultaneously , the order of the approximation the
number of pieces increases.
 In all the cases, the curve is C C2

89
 Computer Aided Design
Splines

3 points, order 3!

90
 Computer Aided Design
Splines

5 points

91
 Computer Aided Design
Splines

11 points

92
 Computer Aided Design
Splines

21 points

93
 Computer Aided Design
Splines


Random perturbation

Each point is moved radially by a value between
-0.5 and +0.5 % of the circle's radius

94
 Computer Aided Design
Splines

3 points, random perturbation 1%

95
 Computer Aided Design
Splines

5 points, random perturbation 1%

96
 Computer Aided Design
Splines

11 points, random perturbation 1%

97
 Computer Aided Design
Splines

21 points, random perturbation 1%

98
 Computer Aided Design
Splines


Deterministic perturbation

Each point is shifted radially depending on its
position by -5 or +5 % of the circle's radius

99
 Computer Aided Design
Splines

3 points, deterministic perturbation 5%

100
 Computer Aided Design
Splines

5 points, deterministic perturbation 5%

101
 Computer Aided Design
Splines

11 points, deterministic perturbation 5%

102
 Computer Aided Design
Splines

21 points, deterministic perturbation 5%

103
 Computer Aided Design
Splines
non local
control

11 points

104
Computer Aided Design
Splines

105
 Computer Aided Design
Splines


Perturbation of a point

We shift one point by a significant amount

106
 Computer Aided Design
Splines

21 points

107
 Computer Aided Design
Splines

99 points

108
 Computer Aided Design
Splines

999 points

109
 Computer Aided Design
Splines


Stable interpolation scheme

Weak Runge phenomenon

The displacement of a point yet affects all the
curve

Nevertheless, the perturbation fades very quickly
further away from the shifted point

« Overshoots » are limited.

110
 Computer Aided Design
Splines


Closed curve ?

The curve can be closed, just impose everywhere
that the second derivative is continuous.

111
 Computer Aided Design
Splines


Instead of
''
x [0] 0=0⇔ 2a[0] 2=0
x ''[n−2] 1=0⇔ 2a[n−2] 26a[ n−2] 3=0
... we have
'' ''
x [n−2] (1)= x (0)⇔ 2a[ n−2 ]2 +6a[ n−2 ]3=2a [0] 2
[0 ]

'
4 1 1 x 3( x1 − x n−2 )0

( )( ) ( )
'
1 4 1 x 3( x 2 − x 0 )1
Circulant 1 4 1 x = 3( x 3 − x 1) '
2 x n−1 = x 0
matrix ⋱ ⋮ ⋮ and ' '
x n−1 = x 0
1 4 1 x 'n−3 3( x n−2− x n−4 )
1 1 4 x' 3( x 0− x n−3 )
n−2 112
 Computer Aided Design
Splines

General case : arbitrary parametrization

x (ui )= xi
P (ui )= P i ≡
{ y (ui )= y i
x[i] (u)= A[i] 0 + A[i ]1 u+ A[i ] 2 u 2 + A[i ]3 u 3 , u∈[u i , u i+1 ]


We again change the parametrization...
u−ui du
̄= 1 = 1
̄u = du u i+1−ui hi
u i +1−u i

2 3
x[i] ( u
̄ )=a[i] 0 + a u
[i]1 ̄ +a u
[i ]2 ̄ + a u
[i] 3 ̄ ,u
̄ ∈[0,1]

113
 Computer Aided Design
Splines


We pass through both points :
2 3
P 
u0 =0= P i ⇔ a [i ]0 a [i]1 
u 0a [i ] 2 
u a [i ]3 
0 u = xi
0
2 3
P 
u1=1= P i1 ⇔ a [i ]0a [i ]1  u1 a [i] 2 
u 1 a [i]3 
u1= x i1

We impose both slopes :
' dP dP 1 2 '
P  u0 =0= 0= 0 = P 'i ⇔ a[i ]12a [i] 2 u0 3a [i] 3 u0 = x i hi
du d u hi
dP dP 1
P '  u1=1= 1= 1 = P ' i1 ⇔a [i ]12a[i ]2 u 13a[i ]3 u 21 = x 'i1 h i
du d u hi
a[i ]0 = x i

{
Finally : a = x ' h

[i ]1 i i

a[i ]2 =3 x i1− xi −2 x 'i h i − x 'i1 h i

' '
a[i ]3=2 x i − x i1  x h i  x
i i1 hi 114
 Computer Aided Design
Splines


We impose the second derivative for a natural
spline 2
d P [i]
2 2
d P [i] d 2 u d P [i] 1
2
 u = 2 2
= 2 2
du d u d u d u hi
'' '' 2a[i−1] 2 6a [i−1] 3 2a [i] 2
x[i−1] 1= x[i ] 0⇔ 2
= 2
hi−1 hi


We substitute in the internal equations
2[3 xi − x i−1 −2 x 'i−1 hi−1 − x 'i hi−1 ]6[ 2 x i−1− x i  x 'i−1 hi−1  x 'i hi−1 ]
2
h i−1

2[3 x i1 − x i −2 x 'i hi − x 'i1 hi ]

= 2
hi 115
 Computer Aided Design
Splines

We obtain finally :

x 'i−1 2 x 'i 2 x 'i  x 'i1  x i − x i−1   x i1 − x i 

 =3 3
hi−1 hi 2
hi−1 hi2

' ' ' ' hi hi−1

hi  x i−1 2 x hi−1 2 x  x
i i i1 =3  x i − x i−1 3  x i1 − x i 
h i−1 hi

116
 Computer Aided Design
Splines


At the boundaries we want a vanishing second
derivative …
'' 2a [0] 2
x (0)=0 ⇔
[0] 2
=0
(h ) 0 h0 (2 x '0 + x '1)=3( x 1− x 0 )
'' 2a[ n−2 ]2 +6a[ n−2 ]3 ' '
h n−2 ( x n−2 +2 x n−1 )=3( x n−1− x n−2 )
x [n−2] (1)=0⇔ 2
=0
(h n−2 )


We have then a linear system of n unknowns:
(next page)
117
 Computer Aided Design
Splines
'
2 h0 h0 x
0

( )( )
'
h1 2 h1 + 2 h 0 h0 x
1
⋱ ⋮ =
h n−2 2 hn−2 + 2 h n−3 h n−3 x 'n−2
h n−2 2 h n−2 x 'n−1

3( x 1 − x 0 )

( )
h1 h0
3 ( x 1− x 0 )+3 ( x 2− x 1 )
h0 h1
⋮
h n−2 h n−3
3 ( x n−2 − x n−3 )+3 ( x n−1 − x n−2 )
h n−3 h n−2
3( x n−1− x n−2 )
118
 Computer Aided Design
Splines
Natural Spline
u=3 (uniform parametrization)

u=4 u=1

u=4
u=5 u=2 u=0

Natural Spline
(non uniform parametrization)

119
 Computer Aided Design
Splines


Compact notation

We would like a compact representation as for the
Lagrange interpolation. a [i]0 = xi

0≤ ū <1

n n
2
x [i ] ( ū)=a [i ]0 +a [i]1 ū+a [i ]2 ū +a [i ]3 ū
3

{ '
a [i]1 =x i
' '
a [i] 2=3 xi1− x i −2 xi −x i1
a [i]3 =2 xi −x i1  xi' x 'i1

{
p ' p
x[i] ( ū)=∑ xi h ( ū)+ ∑ x h ( ū)
i0 i i1
0 0 n n
n n p ' p
p ' p ⇔ P [i ] ( ū)=∑ P i h ( ū)+∑ P h ( ū)
i0 i i1
y[i ] ( ū)=∑ yi h ( ū)+ ∑ y h ( ū)
i0 i i1 0 0
0 0
...
This on each interval [i,i+1].
120
 Computer Aided Design
Splines


We have on each interval :
a [i]0 = xi

{ a [i]1 =x 'i
' '
a [i] 2=3 xi1− x i −2 xi −x i1
a [i]3 =2 xi −x i1  xi' x 'i1

x [i ]  u = xi  x 'i u 3 x i1 −x i −2 x 'i − x'i1  u 22 xi − x i1  x'i  x 'i1  u 3


By rearranging equations
x [i ]  u = xi 1−3 u 2 2 u 3 x 'i  u −2 u 2 u 3 
2 3 ' 2 3
 x i1 3 u −2 u  xi1 −u  u 
121
 Computer Aided Design
Splines


Hermite polynomials (for two points)
p
h 00 =1−3 u 22 u 3

{ p

p
2
h10 =3 u −2 u
p

2
h11=−u  u
2
3

h 01=u −2 u  u
3
3

122
 Computer Aided Design
Splines


The more general Hermite basis of degree 2n-1
allows to do an interpolation of n points on an
interval, as the Lagrange basis, additionally
imposing slopes at each point.
n ''
l (u i )

n
[
hi0n (u)= 1− n'
l (u i )
n 2
]
(u−ui ) [l in (u)]2

h u=u−ui [l u]

i1 i
n n−1
n l u u−ui 
l i u= n ' = ∏
l ui u−ui  j=0,i≠ j u j −u i 
n−1
n
l u=∏ u−u j 
j=0
123
 Computer Aided Design
Splines


Properties of the Hermite basis
Interpolation
of the
hi0n u j = ij positions

h i0n ' u j =0

n
hi1 u j =0
h i1n ' u j =ij
Interpolation
of the slopes

n
∑h i0 u=1
i

124
 Computer Aided Design
Splines


Property of affine invariance

It is a useful property such that the curves we
define for a set of control points can undergo linear
affine transformations without hassle.
 Let Pi* the affine transformation of the control points Pi
 Let P*(u,Pi) the affine transformation of the points of the
curve P(u,Pi) defined from the original points Pi
 Let P(u,Pi*) the new curve based on the modified control
points Pi* , with the same parametrization.
 The affine invariance is verified iff P*(u,Pi) = P(u,Pi*)
for all u.
125
 Computer Aided Design
Splines


Affine transformations  P ≡A⋅P u
Translation a 1 0 0
Scaling
3 rotations
[] [ ]
u= b ; A= 0 1 0
c 0 0 1
d 0 0
Shear

1 g h
[ ]
u=0 ; A= 0 e 0
0 0 f

[ ]
u=0 ; A= 0 1 i
0 0 1
cos  −sin  0
[ ]
u=0 ; A= sin  cos  0 ⋯
0 0 1

12 degrees of freedom

126
 Computer Aided Design
Splines

Let P a parametric curve built this way :

n−1
P (u , P i )=∑ P i K ni (u)
0


Let's verify the invariance by a translation t:
n−1 n−1 n−1
* n n n
P (u , P )=∑ ( P i +t) K (u)=∑ P i K (u)+ ∑ t K (u)
i i i i
0 0 0
n−1
n
= P (u , P i )+ ∑ t K (u) i Partition of unity
0
n−1
= P (u , P i )+t=P * (u , P i ) iff K n
∑ i (u)=1
0 127
 Computer Aided Design
Splines


For the other multiplicative transformations
n−1 n−1
* n n
P (u , P )=∑ (A⋅P i ) K (u)=A⋅∑ P i K (u)
i i i
0 0

=A⋅P (u , P i )=P * (u , P i )
(no particular conditions except linearity
with respect to the coordinates of the control points)
Consequently, iff the basis functions form a partition of unity,
and the dependence with respect to the control points is
linear, then the representation is invariant by any affine
transformation.

128
 Computer Aided Design
Splines


Do Hermite's basis form a partition of unity ?
n−1 p
h 00 =1−3 u 22 u 3

{
n
h
∑ ?? u=1 p 2 3
0 h10 =3 u −2 u
p 2 3
h 01=u −2 u  u
p 2 3
F u=1 h11=−u  u

'
F 1=1 ' F 1=0
F 0=1 F 0=0

129
 Computer Aided Design
Splines


Let's check the invariance
 If we apply a translation to the control points Pi, the
derivatives Pi' should not change ...
* '* '
P = P i +t
i P =P
i i
1 1
P ( P *i )=∑ ( P i +t ) h ni0 (u)+ ∑ P i' hi1n (u)
0 0

1 1 1
n n ' n
=∑ th (u)+ ∑ P i h (u)+ ∑ P h (u)
i0 i0 i i1
0 0 0

*
=t + P (P i )= P ( P i )

130
 Computer Aided Design
Splines

We must also check the invariance for the other
multiplicative transformations : those affect both
the coordinates and the derivatives
* '* '
P i = A⋅P i P i = A⋅P i
1 1
* n ' n
P  P i =∑  A⋅P i h i0 u∑  A⋅P i  hi1 u
0 0

1 1

 n
= A⋅ ∑ P i h u∑ P h u
0
i0
0
'
i
n
i1 
*
= A⋅P ( P i )= P (P i )
QED
131
 Computer Aided Design
Splines
 Beware of the computation of the slopes x'i ...

Natural Splines :
'
x

   
2 1 3 x 1− x0 
0


'
1 4 1 x 3 x 2− x0 
1
'
1 4 1 x = 3 x 3− x1 
2
⋱ ⋮ ⋮
1 4 1 x' 3 x n−1− x n−3 
n−2
1 2 ' 3 x n−1−x n−2 
x n−1
Linear operator

P 'i = L( P i − P j )⇒ P 'i * = L (P *i − P *j )= L(( A⋅P i +t)−( A⋅P j +t))

'
= A⋅L( P i − P j )= A⋅P i It is OK in this case 132
 Computer Aided Design
Splines


Rotation of 45°

Scaling x direction
(times 0.5)

Followed by a
translation
133
 Computer Aided Design
Splines


3D Curves

Minimal order so that a curve can have a torsion
(non planar curve) n−1
p

Let's consider a Lagrange interpolation P u= ∑P i l i u
i=0

2 points → on a straight line (no curvature)

3 points → in a plane (no torsion)

4 points → torsion becomes possible

Minimal order to join smoothly two arbitrarily
oriented curves = 3

134