Introduction to Visualization and Computer Graphics

DH2320, Fall 2015

Prof. Dr. Tino Weinkauf

Geometric Modeling

Introduction
 Geometric Modeling: Introduction

● There are many ways for creating graphical data.

● Classic way: Geometric Modeling

p3

p2
p1

p0

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

● There are many ways for

creating graphical data.

● Other approaches:
● 3D scanners
● Photography for measuring
optical properties
● Simulations, e.g., for flow
data 3D Scanning

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

● Geometric objects convey a part of the real or theoretical

world; often, something tangible

● They are described by their geometric and topological

properties:
● Geometry describes the form and the position/orientation in a
coordinate system.
● Topology defines the fundamental structure that is invariant against
continuous transformations.

Different geometry Different geometry

Same topology Different topology
Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

● Geometric Modeling is the computer-aided design and

manipulation of geometric objects. (CAD)

● It is the basis for:

● computation of geometric properties
● rendering of geometric objects
● physics computations (if some physical attributes are given)

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

● 3D models are geometric representations of 3D objects with

a certain level of abstraction.

● We distinguish between three types of models:

● Wire Frame Models

● describe an object using boundary lines

● Surface Models
● describe an object using boundary surfaces

● Solid Models
● describe an object as a solid

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

Wire Frame Models

● Describe an object using boundary

curves

● No relationship between these

curves
● Surfaces between them are not
defined
non-sense objects (Ernst, 1987)
● Properties:
● simple, traditional
● non-sense objects possible
● visibility of curves cannot be decided
● solid object intersection cannot be
computed
● surfaces between the curves cannot
be computed automatically
● not useable for CAD/CAM
ambiguity of wire frame models

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

Surface Models

● Defines surfaces between boundary

curves

● Describes the hull, but not the

interior of an object

● Often implemented using polygons,

hull of a sphere or ellipsoid, free-
polygonal surface
form surfaces, … representation

● No relationship between the

surfaces
parametric surface
● The interior between them is not representation
defined using 32 Bezier
patches

● Visibility computations: yes

Solid intersection comp.: no
● Most often used type of model

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

Solid Models

● Describe the 3D object

completely by covering the solid

● For every point in 3D, we can

decide whether it is inside or
outside of the solid.

● Visibility and intersection solid model and a cut through it

(Werkbild Strässle, from Ockert, 1993)
computations are fully supported

● Basis for creating solid objects

using computer-aided
manufacturing

visibility computation for lines using a solid model

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
 Geometric Modeling: Introduction

Chrome-cobalt disc with crowns for dental implants, manufactured using WorkNC CAM
Sescoi CAD/CAM
http://www.flickr.com/photos/cadcamzone/4679188766/. Licensed under CC BY-SA 2.0 via Commons -
https://commons.wikimedia.org/wiki/File:Disc_with_dental_implants_made_with_WorkNC.jpg#/media/File:Disc_with_dental_implants_made_with_WorkNC.jpg

Introduction to Visualization and Computer Graphics, Tino Weinkauf, KTH Stockholm, Fall 2015
Introduction to Visualization and Computer Graphics
DH2320, Fall 2015
Prof. Dr. Tino Weinkauf

Geometric Modeling

Bezier Curves, Splines and Surfaces

de Casteljau Algorithm
Bernstein Form
Bezier Splines
Tensor Product Surfaces
Total Degree Surfaces
 Bezier Curves
de Casteljau algorithm

• Paul de Casteljau (1959) @ Citroën

• Pierre Bezier (1963) @ Renault

Meine Zeit bei Citroën / My time at Citroën

see the PDF deCasteljau_de.pdf and deCasteljau_en.pdf in the download area of the webpage
Bezier curves
History:
• Bezier curves/splines developed by
 Paul de Casteljau at Citroën (1959)
 Pierre Bézier at Renault (1963)

for free-form parts in automotive design

• Today: Standard tool for 2D curve editing
• Cubic 2D Bezier curves are everywhere:
 Postscript, PDF, Truetype (quadratic curves), Windows GDI...
 Inkscape, Corel Draw, Adobe Illustrator, Powerpoint, ...

• Widely used in 3D curve & surface modeling as well

All You See is Bezier Curves...
De Casteljau algorithm
Approximation setting:

Given: p0, …, pn

Wanted: smooth, approximating curve

De Casteljau algorithm
Linear interpolation
De Casteljau algorithm
Parabolas

 planar curve, even if defined in R3

Example:
De Casteljau algorithm
Another parabola construction

given: 3 points b0, b1, b2

parabola x(t)
De Casteljau algorithm
Example
De Casteljau algorithm

De Casteljau Algorithm: Computes x (t) for given t

• Bisect control polygon in ratio t : (1 – t)
• Connect the new dots with lines (adjacent segments)
• Interpolate again with the same ratio
• Iterate, until only one point is left
De Casteljau algorithm
Description of the de Casteljau algorithm
• given: points
• wanted: curve

• geometric construction of the point x(t) for given t:

• Then, is the searched curve point x(t) at the

parameter value t
De Casteljau algorithm
repeated convex combination of control points
bi(r) = (1-t)·bi(r-1) + t·bi+1(r-1)
b1(0) b2(0)

b0(0)
b0(0) b3(0)
b1(0)

b2(0)

b3(0)
De Casteljau algorithm
repeated convex combination of control points
bi(r) = (1-t)·bi(r-1) + t·bi+1(r-1)
b1(1)
b1(0) b2(0)

b0(1)
b2(1)
b0(0)
1-t b0(0) b3(0)
t
b1(0) b0(1)
1-t
t
b2(0) b1(1)
1-t
t
b3(0) b2(1)
De Casteljau algorithm
repeated convex combination of control points
bi(r) = (1-t)·bi(r-1) + t·bi+1(r-1)
b1(1)
b1(0) b2(0)
b0(2) b1(2)
b0(1)
b2(1)
b0(0)
1-t b0(0) b3(0)
t
b1(0) b0(1)
1-t 1-t
t t
b2(0) b1(1) b0(2)
1-t 1-t
t t
b3(0) b2(1) b1(2)
De Casteljau algorithm
repeated convex combination of control points
bi(r) = (1-t)·bi(r-1) + t·bi+1(r-1)
b1(1)
b1(0) b2(0)
b0(2) b1(2)
b0(1) b0(3) = x(t)
b2(1)
b0(0)
1-t b0(0) b3(0)
t
b1(0) b0(1)
1-t 1-t
t t
b2(0) b1(1) b0(2)
1-t 1-t 1-t
t t t
b3(0) b2(1) b1(2) b0(3) = x(t)

de Casteljau scheme
De Casteljau algorithm
The intermediate coefficients bir(t) can be written in
a triangular matrix: the de Casteljau scheme:
De Casteljau algorithm
b1(1)
b1(0) b2(0)
b0(2) b1(2)
b0(1) b0(3) = x(t)
b2(1)

Algorithm: b0(0) b3(0)

for r = 1..n do
for i = 0..n-r do
bi(r) = (1-t)·bi(r-1) + t·bi+1(r-1)
end for
end for The whole algorithm consists only of
repeated linear interpolations.
return b0 (n)
De Casteljau algorithm
The polygon consisting of the points b0, …, bn is called
Bezier polygon. The points bi are called Bezier points.

The curve defined by the Bezier points b0, …, bn and the

de Casteljau algorithm is called Bezier curve.

The de Casteljau algorithm is numerically stable, since

only convex combinations are applied.

Complexity of the de Casteljau algorithm

• O(n2) time
• O(n) memory
• with n being the number of Bezier points
De Casteljau algorithm
Properties of Bezier curves:
• given: Bezier points b0, …, bn
Bezier curve x(t)
• Bezier curve is polynomial curve of degree n.
• End point interpolation: x(0) = b0, x(1) = bn. The
remaining Bezier points are only generally approximated.
• Convex hull property:
Bezier curve is completely inside the convex hull of its
Bezier polygon.
De Casteljau algorithm
• Variation diminishing
no line intersects the Bezier curve more often than its Bezier
polygon.

• Influence of Bezier points: global, but pseudo-local

• global: moving a Bezier point changes the whole curve
progression
• pseudo-local: bi has its maximal influence on x(t) at t = i /n.

• Affine invariance:
Bezier curve and Bezier polygon are invariant under affine
transformations

• Invariance under affine parameter transformations

De Casteljau algorithm
• Symmetry:
The following two Bezier curves coincide, they are
only traversed in opposite directions:

• Linear precision:
Bezier curve is line segment, if b0,…, bn are
collinear

• Invariant under barycentric combinations

De Casteljau algorithm
• First derivative of a Bezier curve
• Endpoints:

b0(0)
1-t b1(1)
b1(0) b2(0)
b0(2) b1(2)
t
b1(0) b0(1) b0(1) b0(3) = x(t)
1-t 1-t b2(1)
t t
b2(0) b1(1) b0(2) b0(0) b3(0)
1-t 1-t 1-t
-1
1 𝑛−1 𝑛−1
b3(0) t b2(1) t
b1(2)
t 𝐱 𝑡 = 𝑛 𝐛1 − 𝐛0

de Casteljau scheme
De Casteljau algorithm
• Second derivative of a Bezier curve

b0(0)
1-t b1(1)
b1(0) b2(0)
b0(2) b1(2)
t
b1(0) b0(1) b0(1) b0(3) = x(t)
1-t -1 b2(1)
t 1
b2(0) b1(1) b0(2) b0(0) b3(0)
1-t -1 -1
t 1 1 𝑛−2 𝑛−2 𝑛−2
b3(0) b2(1) b1(2) 𝐱 𝑡 =𝑛 𝑛−1 𝐛2 − 2𝐛1 + 𝐛0

de Casteljau scheme
Bezier Curves
Bernstein form
Bernstein Basis
Bezier curves are algebraically defined using the
Bernstein basis:
• Bernstein basis of degree n: 
B  B0( n) , B1( n) ,..., Bn( n) 
 n i
B (t ) :   t 1  t 
( n) ni
i
i

n = 2 (quad.) n = 3 (cubic) n = 10
B0 B2 B0 B3
B1 B1 B2
Bernstein Basis

de Casteljau algorithm

𝑛
𝑛 Bernstein form
𝑓 𝑡 = 𝐵𝑖 𝑡 𝐩i
𝑖=0

curve basis function control point

Examples
The first three Bernstein bases:
n = 1 (linear)
B0(0) : 1
B0 B1
B0(1) : 1  t  B1(1) : t

B0(2) : 1  t  B1(2) : 2t 1  t 
2
B2(2) : t 2

B0(3) : 1  t  B1(3) : 3t 1  t 
3 2

n = 2 (quad.) n = 3 (cubic)
B2(3) : 3t 2 1  t  B3(3) : t 3 B0 B2 B0 B3
B1 B1 B2

 n i
B (t ) :   t 1  t 
( n) ni
i
i
Bezier Curves in Bernstein form
Bezier Curves: p2 n = 3 (cubic)
n
• f (t )   pi Bi( n) p1
B0 B3
i 0
p3 B1 B2
t  [0..1]

p0

p3
p3

p0 p2
p1
p2
p1 p3
p2
p1
p0
p0
Summary for Bezier Curves
Bezier curves and curve design:
• The rough form is specified by the position of the control
points
• Result: smooth curve approximating the control points
• Computation / Representation:
 de Casteljau algorithm
 Bernstein form

• Problems:
 high polynomial degree
 moving a control point can change the whole curve
 interpolation of points
  Bezier splines
Towards Bezier Splines

Approximation Interpolation
Towards Bezier Splines
Interpolation problem:
• given:
control points
knot sequence

• wanted:
interpolating curve x(t), i.e., x(ti) = ki for i = 0, …, n

• Approach:
"Joining" of n Bezier curves with certain intersection conditions
Towards Bezier Splines
The following issues arise when stitching together
Bezier curves:
• Continuity
• Degree
• (Parameterization)
 Bezier Splines
Parametric and Geometric Continuity
Continuity
Joining of curves - continuity
• given: 2 curves
x1(t) over [t0, t1]
x2(t) over [t1, t2]

• x1 and x2 are Cr continuous in t1, if they coincide in

0th – rth derivative vector in t1.
Continuity

C-1 continuity C0 continuity

C1 continuity C2 continuity
Continuity
Parametric Continuity Cr:
• C0, C1, C2... continuity.
• Does a particle moving on this curve have a smooth
trajectory (position, velocity, acceleration,...)?
• Useful for animation (object movement, camera paths)
• Depends on parameterization

Geometric Continuity Gr:

• Independent of parameterization
• Is the curve itself smooth?
• More relevant for modeling (curve design)
Bezier Splines
Local control: Bezier splines
• Concatenate several curve segments
• Question: Which constraints to place upon the control
points in order to get C-1, C0, C1, C2 continuity?

p(i)
2
p(i+1)
p1(i) 3
p(i)
3

p(i+1)
p0(i) 0
p(i+1)
2
p1(i+1)
Bezier Spline Continuity
Rules for Bezier spline continuity:
• C0 continuity:
 Each spline segment interpolates the first and last control point
 Therefore: Points of neighboring segments have to coincide for
C0 continuity.

p(i+1)
2
p1(i+1)
p(i)
2
p(i+1)
p1(i) 3
p(i)
3

p(i+1)
0
p0(i)
Bezier Spline Continuity
Rules for Bezier spline continuity:
• Additional requirement for C1 continuity:
 Tangent vectors are proportional to differences p1 – p0, pn – pn-1
 Therefore: These vectors must be identical for C1 continuity

p(i+1)
2
p(i)
2
p(i+1)
p1(i) 3
p(i)
3

p0(i) p(i+1)
0
p1(i+1)
Bezier Spline Continuity
Rules for Bezier spline continuity:
• Additional requirement for C2 continuity:
 𝐝 − = 𝐝+
Continuity

C-1 continuity C0 continuity

G1 continuity C1 continuity
 Bezier Splines
Choosing the degree
Choosing the Degree...
Candidates:

• d = 0 (piecewise constant): not smooth

• d = 1 (piecewise linear): not smooth enough

• d = 2 (piecewise quadratic): constant 2nd

derivative, still too inflexible

• d = 3 (piecewise cubic): degree of choice

for computer graphics applications
Cubic Splines
Cubic piecewise polynomials:
• We can attain C2 continuity without fixing the second
derivative throughout the curve
• C2 continuity is perceptually important
 We can see second order shading discontinuities
(esp.: reflective objects)
 Motion: continuous position, velocity & acceleration
Discontinuous acceleration noticeable (object/camera motion)
• One more argument for cubics:
 Among all C2 curves that interpolate a set of points (and obey to
the same end conditions), a piecewise cubic curve has the least
integral acceleration (“smoothest curve you can get”).
– see AdditionalMaterial/CubicsMinimizeAcceleration.pdf
Spline Surfaces
Spline Surfaces
Two different approaches
• Tensor product surfaces
 Simple construction
 Everything carries over
from curve case
 Quad patches
 Degree anisotropy

• Total degree surfaces

 Not as straightforward
 Isotropic degree
 Triangle patches
 “Natural” generalization of curves
Tensor Product Surfaces
Tensor Product Bezier Surfaces
Bezier curves: b01 b11
repeated linear interpolation
now a different setup: x(u,v)
4 points b00, b10, b11, b01
parameter area [0,1]  [0,1]
b00 b10
bilinear interpolation:
repeated linear interpolation

repeated bilinear interpolation:

gives us tensor product Bezier surfaces
(example shows quadratic Bezier surface)
 De Casteljau Algorithm
De Casteljau algorithm for tensor product surfaces:
b(0,0,0; 1,1,1)
b(1,1,1; 1,1,1) b(u,1,1; v,1,1)
b(u,0,0; v,1,1)
b(u,u,1; v,v,1)
b(u,u,0; v,v,1)

v b(u,u,u; v,v,v) b(u,u,1; v,v,0)

b(1,1,1; 0,0,0)
b(u,u,0; v,v,0) b(u,1,1; v,0,0)

b(u,0,0; v,0,0)
b(0,0,0; 0,0,0) u
Tensor Product Surfaces
Tensor Product Surfaces:
n n
f (u, v )   bi (u)b j (v )pi , j
i 1 j 1
n n
  bi (u) b j (v )pi , j
i 1 j 1
n n
  b j (u) bi (v )pi , j
j 1 i 1

• “Curves of Curves”
• Order does not matter
Tensor Product Surfaces
Bezier Patches
Bezier Patches
Bezier Patches:
• Remember endpoint interpolation:
 Boundary curves are Bezier
curves of the boundary
control points

d=3 d = 10
B0 B3
B1 B2
Continuity Conditions
For C0 continuity:
• Boundary control points must match

For C1 continuity:
• Difference vectors must match at the boundary
C0 Continuity
C1 Continuity
C1 Continuity
Total Degree Surfaces
Spline Surfaces
Two different approaches
• Tensor product surfaces
 Simple construction
 Everything carries over
from curve case
 Quad patches
 Degree anisotropy

• Total degree surfaces

 Not as straightforward
 Isotropic degree
 Triangle patches
 “Natural” generalization of curves
Bezier Triangles
Alternative surface definition: Bezier triangles
• Constructed according to given
total degree
 Completely symmetric:
No degree anisotropy
• Can be derived using a triangular
de Casteljau algorithm
 Barycentric interpolation
Barycentric Coordinates
Barycentric Coordinates: p2
• Planar case:
Barycentric combinations of 3 points  p 
p  p1   p 2   p3 , with :       1

 1     p1
p3
• Area formulation:
area((p 2 , p3 , p)) area((p1, p3 , p)) area((p1, p 2 , p))
 ,  , 
area((p1, p 2 , p3 )) area((p1, p 2 , p3 )) area((p1, p 2 , p3 ))
Example
Cubic Bezier Triangle: c
p(c,c,c)

p(c,c,a) p(c,c,b)

p(c,a,a) p(c,b,b)
p(a,b,c)

a b
p(a,a,a) p(b,a,a) p(b,b,a) p(b,b,b)
De Casteljau Algorithm
c
x   a   b   c, p(c,c,c)
    1
p(x,c,c)

p(c,c,a)
p(c,c,b)
p(x,x,c)

a p(a,a,a) p(c,a,a) p(x,a,c)

p(x,x,x) p(x,b,c)
p(x,a,a)
p(x,x,a) p(a,b,c)
p(c,b,b)
p(x,x,b)
p(x,a,b)
p(b,a,a) p(x,b,b)
p(b,b,b)
p(b,b,a) b
Continuity
We need to assemble Bezier triangles continuously:
• What are the conditions for C0, C1 continuity?
• As an example, we will look at the quadratic case...
Continuity
Situation: b

G d
F
a
c
• Two Bezier triangles meet along a common edge.
 Parametrization: T1 = {a, b, c}, T2 = {c, b, d}
 Polynomial surfaces F(T1), G(T2)
 Control points:
– F(T1): f(a, a), f(a, b), f(b, b), f(a, c), f(c, c), f(b, c)
– G(T2): g(d, d), g(d, b), g(b, b), g(d, c), g(c, c), g(b, c)
Continuity
Situation:
f(b, b) b g(b, b)
g(b, d)
f(a, b)
d
G g(d, d)
f(b, c) g(b, c)

a F
g(c, d)
f(a, a)
f(a, c)
f(c, c) c g(c, c)
Continuity
C0 Continuity:
• The points on the boundary
have to agree:
f(b, b) = g(b, b)
f(b, c) = g(b, c)
f(c, c) = g(c, c)
• Proof: Let x :  b   c,     1
f ( x , x )   f ( b , x )   f (c , x )
  2 f ( b, b) 2 f ( b, c)  2 f (c , c )
|| || ||
g( b,b) g( b,c ) g(c ,c )
  2 g( b, b)  2 g( b, c)   2 g(c , c )
  g( b, x )   g(c , x )  g( x , x )
Continuity
C1 Continuity:
• We need C0 continuity.
In addition:
• Points at hatched
quadrilaterals are coplanar
• Hatched quadrilaterals are
an affine image of the same
parameter quadrilateral
Curves on Surfaces, trimmed NURBS
Quad patch problem:
• All of our shapes are parameterized over rectangular or
triangular regions
• General boundary curves are hard to create
• Topology fixed to a disc (or cylinder, torus)
• No holes in the middle
• Assembling complicated shapes is painful
 Lots of pieces
 Continuity conditions for assembling pieces become complicated
 Cannot use C2 B-Splines continuity along boundaries when using
multiple pieces
Curves on Surfaces, trimmed NURBS
Consequence:
• We need more control over the parameter domain
• One solution is trimming using curves on surfaces (CONS)
• Standard tool in CAD: trimmed NURBS

Basic idea:
• Specify a curve in the parameter domain that
encapsulates one (or more) pieces of area
• Tessellate the parameter domain accordingly to cut out
the trimmed piece (rendering)
Curves-on-Surfaces (CONS)
Curves-on-Surfaces (CONS)
Curves-on-Surfaces (CONS)
Summary
• Bezier Curves
• de Casteljau algorithm
• Bernstein form

• Bezier Splines

• Bezier Tensor Product Surfaces

• Bezier Total Degree Surfaces