Implicit Surface Modeling

Hongxin Zhang and Jieqing Feng

2007-01-04

State Key Lab of CAD&CG

Zhejiang University
 Contents

Introduction of Implicit Surface Modeling

Implicit Surface Modeling for Computer
Graphics and Animation: Field Function
Defined Implicit Surface
Quadric Surfaces
Download Course

01/04/2007 State Key Lab of CAD&CG 2

 Introduction of Implicit Surface
Modeling

What is implicit surface?

Comparison: implicit v.s. parametric
surfaces
Implicit methods for graphics and
animation
Implicit methods for CAD/CAM
Other topics in implicit modeling

01/04/2007 State Key Lab of CAD&CG 3

 What is implicit surface?

Implicit surfaces are two-dimensional,

geometric shapes that exist in three
dimensional space.
Defined in R3:
2-D manifold:
A surface embedded in R3
Infinitesimal neighborhood around any point on the
surface is topologically equivalent ( locally
diffeomorphic) to a disk.

01/04/2007 State Key Lab of CAD&CG 4

 What is implicit surface?

2D manifold can be
bounded (or closed): sphere
Unbounded: plane
A manifold with boundary: topologically
equivalent to either a disk or a half-disk
locally.
Nonmanifold

Manifold Manifold with boundary Nonmanifold

01/04/2007 State Key Lab of CAD&CG 5

 What is implicit surface?

Manifold or not? Eule-Poincar formula

v-e+f=2-2h
v: number of vertices
e: number of edges
f: number of faces
h: number of holes

01/04/2007 State Key Lab of CAD&CG 6

 Examples of implicit surfaces

Metaball Skeleton Surface Convolution

and Metaball Surface

01/04/2007 State Key Lab of CAD&CG 7

 Examples of implicit surfaces

Quadric Surface A-Patch Compactly

Supported Radial
Basis Functions

01/04/2007 State Key Lab of CAD&CG 8

 Definition of implicit surface

Definition
{p=(x,y,z): f(p)=0, pR3}
Implicit function f : three classes of
specification (definition)
Discrete sampling: a set of points on or within an object
Mathematical function: one or more equations are used to
compute the coordinates of points on or within an object
Procedural methods: an algorithmic process computes
points on or within an object

01/04/2007 State Key Lab of CAD&CG 9

 Algebraic Function

Algebraic surface: when f is algebraic

function, i.e., polynomial function
The coefficients are not unique
f=ax+by+cz=0
a=b=c=1
a=b=c=1/3
Algebraic distance: The value of f(p) is the
approximation of distance from p to the
algebraic surface f=0
01/04/2007 State Key Lab of CAD&CG 10
 Transcendental Function

Transcendental Function
Trigonometric, exponential, logarithmic,
hyperbolic functions, etc.
Approximated by convergent power series:
Taylor series

01/04/2007 State Key Lab of CAD&CG 11

 Definition of implicit surface

Regular point p on the surface

f (p) =(f/x, f/y, f/z) 0
For cone, 0 is not the regular value for f

The cone f=-x2+y2+z2 is regular with the exception of a

singularity at the origin (0,0,0).

01/04/2007 State Key Lab of CAD&CG 12

 Mathematical Foundation

Implicit Function Theorem:

if 0 a regular value of continuous function f(p),
the implicit surface f-1(0) is a two-dimensional
manifold
Jordan-Brouwer Separation Theorem
A 2D manifold separates R3 into surface itself
and two connected open sets: an infinite(finite)
`outside' and a finite(infinite) inside'.

01/04/2007 State Key Lab of CAD&CG 13

 Mathematical Foundation

f > 0 outside

f < 0 inside f = 0 surface

f (x,y,z) -1=x2+y2+z2-1 f (x,y,z) = x2+y2+z2

01/04/2007 State Key Lab of CAD&CG 14

 Implicit v.s. Parametric Surfaces

01/04/2007 State Key Lab of CAD&CG 15

 Complex Surface Modeling

Complex surface constructed by

smoothly joined B-spline surface
patches Complex surface constructed by
convolution surface method

01/04/2007 State Key Lab of CAD&CG 16

 Surfaces Intersection

Parametric Surface Intersection Implicit Surface Intersection

01/04/2007 State Key Lab of CAD&CG 17

 Tessellation

Up: tessellation of implicit surface

Bottom: tessellation of NURBS

surface

01/04/2007 State Key Lab of CAD&CG 18

 Blending or rounding

Left: Blending by using parametric surface

Right: Blending by using implicit surface

01/04/2007 State Key Lab of CAD&CG 19

 Texture mapping

Texture mapping for NURBS surface Texture mapping for implicit surface

01/04/2007 State Key Lab of CAD&CG 20

 Solid Modeling

Solid Modeling via NURBS Surface Solid Modeling via Implicit Surface

01/04/2007 State Key Lab of CAD&CG 21

 Implicit v.s. Parametric Surfaces

Implicitization:
Parametric Implicit
representation representation

Parameterization
Implicit Parametric
representation representation

01/04/2007 State Key Lab of CAD&CG 22

 About conversions

Implicitization
Any rational parametric surface can be implictized by
elimination of parameters in the parametric form
Implicitization is often computational demanding
The degree of implicit form is higher than its
parametric form, the implicit representation of
A parametric triangular patch of degree n is degree n2
A tensor product surface of degree m by n is degree 2mn
The number of terms in algebraic surface of degree n is 3C n2
(bicubic patch is degree 18, with 1330 terms!)
Approximated implicitization

01/04/2007 State Key Lab of CAD&CG 23

 About conversions

Parameterization: not always possible

Algebraic surfaces of fourth and higher
degree cannot be parameterized by rational
functions
Parameterization is always possible for non-
degenerate quadrics and for cubics that have
a singular point.

01/04/2007 State Key Lab of CAD&CG 24

 Example:sphere

Trigonometric:
f(,)=(coscos, cossin, sin)
[0,], [0,2]
Rational:
x=4st, y=2t(1-s2), z=(1-t2)(1+s2), w=(1+t2)(1+s2)

Implicit:
f(x,y,z)=x2+y2+z2-1

01/04/2007 State Key Lab of CAD&CG 25

 Example: sphere

Points on the parametrically defined

sphere are readily found by substitution of
and into the equations for x, y, and z
(similarly for s and t).
The mapping from parametric space to
geometric space is convenient.
There is no obvious mapping for implicit
form!

01/04/2007 State Key Lab of CAD&CG 26

 Implicit methods for graphics and
animation

Blobby (metaball, soft objects)

Implicit surface defined by skeletons
Distance surface
Convolution surface
Variational Implicit Surfaces
Level-Set Methods
Procedural Models
Animation applications

01/04/2007 State Key Lab of CAD&CG 27

 Blobby (metaball, soft objects)

01/04/2007 State Key Lab of CAD&CG 28

 Implicit surface defined by
skeletons

01/04/2007 State Key Lab of CAD&CG 29

 Variational Implicit Surfaces

01/04/2007 State Key Lab of CAD&CG 30

 Level-Set Surface Models

01/04/2007 State Key Lab of CAD&CG 31

 Procedural Models

01/04/2007 State Key Lab of CAD&CG 32

 Reconstruction

The polygonal
bunny model was
approximated by
metaballs.
01/04/2007 State Key Lab of CAD&CG 33
 Animation applications

01/04/2007 State Key Lab of CAD&CG 34

 Implicit methods for CAD/CAM

Quadric surface
Algebraic surface patches
A-patch
Tensor-product algebraic surface patches

01/04/2007 State Key Lab of CAD&CG 35

 Quadric Surface

01/04/2007 State Key Lab of CAD&CG 36

 A-patch: Algebraic Surface Patch

01/04/2007 State Key Lab of CAD&CG 37

 A-patch: Algebraic Surface Patch

01/04/2007 State Key Lab of CAD&CG 38

 Tensor-product algebraic surface
patches

01/04/2007 State Key Lab of CAD&CG 39

 Tensor-product algebraic surface
patches

01/04/2007 State Key Lab of CAD&CG 40

 Other Topics in Implicit
Modeling

Functional representation
Visualizing implicit surfaces
Animation by using implicit models
Texture mapping

01/04/2007 State Key Lab of CAD&CG 41

 Functional Representation (F-
rep)

F-rep: denfines a whole geometric object

by a single real continuous function of
several variables as F(X)0
At least C0 continuous
F: a formula or an evaluation procedure
F-rep: combines classic implicits, skeleton
based implicits, set-theoretic solids, sweeps,
volumetric objects, parametric and procedural
models
01/04/2007 State Key Lab of CAD&CG 42
 Functional Representation (F-
rep)
Set-theoretic operations Relations
Blending set-theoretic Sweeping by a moving
operations solid
Offsetting Deformation with
Cartesian product algebraic sums
Bijective mapping Three-dimensional
Metamorphosis texture modeling
Interaction

01/04/2007 State Key Lab of CAD&CG 43

 F-rep: examples

Moving solid

Artistic shape modeling using real functions

01/04/2007 State Key Lab of CAD&CG 44
 F-rep: hairs

Using real functions with application to hair modeling

01/04/2007 State Key Lab of CAD&CG 45
 Visualizing implicit surfaces

Polygonization: Generation of polygons

from implicit surface or volume data
Uniform
Adaptive
Ray tracing
Classical
Sphere tracing
Particle system

01/04/2007 State Key Lab of CAD&CG 46

 Uniform Polygonization

A sphere approximated from depths 1 through 5.

01/04/2007 State Key Lab of CAD&CG 47
 Adaptive Polygonization

01/04/2007 State Key Lab of CAD&CG 48

 Polygonization of non-manifold

01/04/2007 State Key Lab of CAD&CG 49

 Classic Ray Tracing

01/04/2007 State Key Lab of CAD&CG 50

 Sphere tracing

01/04/2007 State Key Lab of CAD&CG 51

 Sphere tracing

01/04/2007 State Key Lab of CAD&CG 52

 Particle system: display and control

[Link]
01/04/2007 State Key Lab of CAD&CG 53
 Field Function Defined Implicit
Surface

Field Function in 3D
Simple Control Primitives: point etc.
Blobby Molecules
Metaball
Soft Objects
Discussion and Extensions
Skeletal Primitives
Distance surface
Convolution surface
01/04/2007 State Key Lab of CAD&CG 54
 Field Function in 3D

Consider implicit surface

D(x,y,z)-Iso=0
D is called scalar field function
Distance field is one of good choice

Distance to a Curve: 3 and 9 linear segments

01/04/2007 State Key Lab of CAD&CG 55

 Example of Distance Field

Consider function D(r)=1/r2 , control points in 3D space,

where r is the distance to one control points
The total field strength at any point in space is the sum
of the field strengths due to each control point

01/04/2007 State Key Lab of CAD&CG 56

 Example of Distance Field

Changing control points as line segments

Simple summation
of distance fields
will result in bulge!

01/04/2007 State Key Lab of CAD&CG 57

 Blobby Molecules by Blinn

Second Graphics Fellow at

Microsoft
1983 -- The first Siggraph
Computer Graphics Achievement
Award for work in lighting and
surface modeling techniques.
1989 -- IEEE Outstanding
Contribution Award for Jim Blinn's
Jim Blinn corner
[Link]

01/04/2007 State Key Lab of CAD&CG 58

 Blobby Molecules

Hydrogen atoms electron density fields

(Gaussian Distribution)
br 2
D ( r ) = ae
"b" : standard deviation of Gaussian curve
"a" : height of Gaussian Curve
r : the distance to atom center, r0

2D Gaussian Function 1D Gaussian Function

01/04/2007 State Key Lab of CAD&CG 59

 Multiple Blobby Molecules

The overall density at a given point for multiple

atoms is summation of individual ones:

D ( p ) = ai e bi r 2
= Iso
i
The blobbiness of a model can be controlled by
adjusting the parameters ai and bi.
ai is weight/strength of ith blobby molecule
bi is influence radius of ith blobby molecule
The implicit surface is defined as all points where the
density is equal to some threshold value Iso.

01/04/2007 State Key Lab of CAD&CG 60

 Examples of Blobby Models

01/04/2007 State Key Lab of CAD&CG 61

 Blobby Molecules

Disadvantages of Blobby Molecules

Exponential distribution of density is Global
Computationally expensive!
Blinn: Cut off after r >R
R

01/04/2007 State Key Lab of CAD&CG 62

 Metaball by Nishimura

Nishimura at Osaka University in

Japan proposed Metaball in 1983
(Almost same with Blinn)

01/04/2007 State Key Lab of CAD&CG 63

 Metaball by Nishimura

Piecewise quadrics to approximate the Gaussian

function
3r 2
a 1 2 0 r b/3
b
3a 2
r
D(r ) = 1 b/3 r b
2 b
0 r b

"b" : Influence radius of metaball
"a" : weight of metaball
r : the distance to metaball center, r0
01/04/2007 State Key Lab of CAD&CG 64
 Metaball Systems

The system is composed of several

metaball

D ( p ) = Di ( p ) = Iso
i

Each metaball has its own parameters (ai, bi)

Fusion effects

01/04/2007 State Key Lab of CAD&CG 65

 Examples of Metaballs

01/04/2007 State Key Lab of CAD&CG 66

 Metaball

Advantages
Field function is locally defined in [0,b]
Fast ray and metaball intersection
computation!
Disadvantage
Distance r: square root , expensive

01/04/2007 State Key Lab of CAD&CG 67

 Soft Object by Wyvill Brothers

Soft objects is proposed by Wyvills in Canada

and New Zealand: truncated Taylor expansion
series of exponential function

Brian Wyvill Geoff Wyvill

01/04/2007 State Key Lab of CAD&CG 68

 Field Function for Soft Object

Definition

4r6 17r4 22r2

a1 6 + 4 2 r b
D(r) = 9b 9b 9b
0 r >b

The function has following properties
D(0)=1, D(b)=0, D(b/2)=0.5, C(0)=C(b)=0
The volume bounded by D(r)<m is one half of that
bounded by D(r)<m/2
01/04/2007 State Key Lab of CAD&CG 69
 Examples of Soft Objects

01/04/2007 State Key Lab of CAD&CG 70

 Examples of Soft Objects

Two blend soft objects with different threshold: 84, 44, 24, 11

25 balls, threshold is 0.5 Trains modeled by soft objects

01/04/2007 State Key Lab of CAD&CG 71

 Soft Object

Advantages
Only square term of the distance r, square
root computation is removed!
Finite extent of each ball
Computational costs:
Blobby >> Metaball > Soft object

01/04/2007 State Key Lab of CAD&CG 72

 Discussions

The functions described here guarantee a continuously

and smoothly changing surface
Comparison of four field functions

rt rmax

The blending function can be designed freely!

01/04/2007 State Key Lab of CAD&CG 73

 Extensions

Minus primitives: set weight negative

01/04/2007 State Key Lab of CAD&CG 74

 Applications of Metaball

Shape approximation and representation by metaballs

01/04/2007 State Key Lab of CAD&CG 75
 Skeletal Primitives

Disadvantage of point primitives: flat

surfaces can only be approximated.

Evenly spaced grid of metaballs(16): threshold is 1.66 or 0.5

01/04/2007 State Key Lab of CAD&CG 76

 Skeletal Primitives

The use of complex skeletons is one possible

solution to this problem.
Point can be replace as lines, polygons, curves,
surfaces , volumes and any other complex geometric
skeletons
The shape of a primitive follows the shape of its
skeletons
Two Approaches
Distance surfaces
Convolution Surfaces

01/04/2007 State Key Lab of CAD&CG 77

 Examples of Skeletal Primitives

skeletons

skeletons

contour line drawing

ray traced image

01/04/2007 State Key Lab of CAD&CG 78
 Examples of Skeletal Primitives

Complex skeletons: ellipsoid, sphere, line segments, points

01/04/2007 State Key Lab of CAD&CG 79

 Distance surface

A distance surface is a surface that is

defined by distance to some set of base
skeleton elements
Union of skeletons: Distance surface to the union of
skeletons
Union of primitive surfaces: Boolean operation
Blending of primitive surfaces: Algebraic blending

01/04/2007 State Key Lab of CAD&CG 80

 Union of Skeletons

Definition
s p 2

f ( S , p ) = max sS exp
2

S is the skeleton set
s is the point on one of the skeleton
max: maximum field value, nearest point on
the skeleton
Field function can be specified as before,
not restricted as exponential function
01/04/2007 State Key Lab of CAD&CG 81
 About Union of Skeletons

Distance surface is rounded when skeleton is convex.

Distance surface is tangent discontinuous and exhibits a
crease when the skeleton is concave

Left: curve shown as dashed and distance shown as greyscale intensity

middle and right: curve approximated by three and nine segments

01/04/2007 State Key Lab of CAD&CG 82

 Union of Distance Surfaces

Union of distance surface is defined as

weighted union of each primitive surface
Definition
Ftotal ( p ) = i ci Fi ( ri )

ci: weight, positive or negative

Fi: field function as before
ri: the distance from p to the nearest point on
the i-th skeleton.
01/04/2007 State Key Lab of CAD&CG 83
 About Union of Distance Surfaces

Simple union of distance surfaces will

produce bulges and creases may area
where the skeletons meet.

01/04/2007 State Key Lab of CAD&CG 84

 Algebraic Blends of Primitives

To eliminate creases or bulge , the primitive

volumes (metaball ) must form a blend, rather
than a union.
What is blending: smooth transition along
common boundary among several primitives

01/04/2007 State Key Lab of CAD&CG 85

 Blending Functions (1)

Super-elliptical Blending

P1, P2 are algebraic distances to skeletal

elements 1 and 2, usually C1 continuous
r1 and r2 are the ranges of influence for
primitives P1 and P2
[x]+ is max (0, x), and t is the thumbweight.

01/04/2007 State Key Lab of CAD&CG 86

 Example of Blending
Functions (1)

Super-elliptical Blending of sphere and cylinder primitives (t=3)

01/04/2007 State Key Lab of CAD&CG 87

 Example of Blending
Functions (1)

Super-Elliptic Blend of Two Cylinders

There is a bulge where the cylinders intersect.

01/04/2007 State Key Lab of CAD&CG 88

 Blending Functions (2)

Improved Super-elliptical Blending

: the angle between the gradients of the two

primitives at a point p:
The influence range is diminished according
to

01/04/2007 State Key Lab of CAD&CG 89

 Blending Functions (2)

= 0 the range is fully diminished and the simple

union of the primitives results
= 90(concave condition), the range is undiminished,
and a blend occurs.
To avoid enlarging the primitive ranges, cos() must
be nonnegative

simple union bulging blend use of cos use of nonnegative cos

01/04/2007 State Key Lab of CAD&CG 90

 Blending Functions (3)

Super-elliptical blend is extended to k

primitives:

01/04/2007 State Key Lab of CAD&CG 91

 Convolution Surface

A bulge-free implicit blend technique

The convolution surface treats a skeleton

S as a set of points, each of which
contributes to the implicit surface function
according to its distance to p.

01/04/2007 State Key Lab of CAD&CG 92

 Convolution Surface

Blobby systems by Blinn

where si is a point on the skeleton, c is threshold

(constant)

01/04/2007 State Key Lab of CAD&CG 93

 Convolution Surface
If S = {si} is a set of infinitesimally spaced
points, f can be expressed as an integral:

where u ranges over all points on the

skeleton. It defines a convolution surface
Gaussian function is a 3D filter/kernel
S = {si} can be any skeletons: point, line,
curve, surface, volumes etc.
01/04/2007 State Key Lab of CAD&CG 94
 About Convolution Surface

The convolution has been applied

extensively in the processing of 1D audio
signal, 2D image, 3D volume image.
The convolution scaled the frequency
components of the signal via filter(kernel)
Low-pass filter: smoothing effect, e.g.,
Gaussian kernel
High-pass filter: enhance detail effect

01/04/2007 State Key Lab of CAD&CG 95

 Gaussian Kernels

Gaussian Kernels: left: one-dimensional, right: two-dimensional

Infinite Sum is a Convolution

01/04/2007 State Key Lab of CAD&CG 96

 About Convolution Surface

The convolution is a linear operator:

h ( s1 + s2 ) = h s1 + h s2
It is property of superposition
The sum of the convolutions of any division of
a skeleton is identically equal to the single
convolution of the entire skeleton
The convolution surface is a smooth
shape without introducing bulges.
01/04/2007 State Key Lab of CAD&CG 97
 Illustration of Superposition

The superposition Property: The two line segment are brought

together (left); sum of convolutions of the two segments (right),
convolution of single segment (bottom)

Two Segments Convolved with the Gaussian Kernel

01/04/2007 State Key Lab of CAD&CG 98

 About Convolution Surface

Along convex portions of the skeleton the surface

mimics the union operator
Along concave portions, the surface yields a blend.
For isolated convex skeletons, such as triangles or
segments, convolution produces surfaces of similar
shape to distance surfaces.
For complex skeletons, however, convolution yields
crease-free surfaces with adjacent primitives blending
without seam or bulge.

01/04/2007 State Key Lab of CAD&CG 99

 Gaussian kernel

For standard Gaussian kernel:

x2
1

2

e 2
dx = 1

A signal modified by such a kernel will

maintain its original energy!
1

1/2

Convolution of a box with unit-integral kernel

01/04/2007 State Key Lab of CAD&CG 100

 Energy Discussion in Convolution

Because the Gaussian is symmetric, the

convolution equals where the box function
undergoes transitions
For an iso-surface contour level of , the convolution
surface will pass through the endpoints of skeletal
segments and through the edges of skeletal polygons.

A kernel with integral less (greater) than one

would attenuate (amplify) a signal

01/04/2007 State Key Lab of CAD&CG 101

 Gaussian kernel

Advatnages of Gaussian Kernel

C Smooth
Unit integral
Separable

Disadvantages of Gaussian Kernel

Globally Defined on the R1,2,3
Cannot be analytical integrated, only numerically
approximated
01/04/2007 State Key Lab of CAD&CG 102
 Examples of Convolution Surfaces

Two Segments Convolved with the Gaussian Kernel

01/04/2007 State Key Lab of CAD&CG 103

 Examples of Convolution Surfaces

Threshold=1.0, 0.75, 0.5, 0.35

Line segment and arc skeletons

01/04/2007 State Key Lab of CAD&CG 104

 Examples of Convolution Surfaces

Convolution Surface

Convolution Surface: Hand Crafted by Jules Bloomenthal

01/04/2007 State Key Lab of CAD&CG 105
 Other kernels: Sinc Kernel

Sinc kernel is : sin(x)/x.

The Sinc Kernel

The sinc is the ideal low-pass filter: its Fourier

transform is the box function.
Cutoff frequency effect

01/04/2007 State Key Lab of CAD&CG 106

 Other kernels: Sinc Kernel

Problem with sinc kernel: none monotonically

Convolution with Gaussian (left) and Sinc (right)

Monotonicity of the kernel is a necessary

property for a satisfactory convolution surface
Ringing of the filter may introduce a spurious contour
01/04/2007 State Key Lab of CAD&CG 107
 Other kernels: B-spline Kernels

B-spline kernel has similar shape with

Gaussian

hB-spline(0)=2/3, hB-spline(0.72235)1/2
The support of kernel is 2
The integral of B-Spline kernel is 1, not separable

01/04/2007 State Key Lab of CAD&CG 108

 Other kernels: Wyvill Kernels

Wyvill kernel also has similar shape with

Gaussian
hWyvill(x) = (9-4x6+17x4-22x2)/9.
After scale: hWyvill(0)=1, hWyvill(1)=1/2
The integral of Wyvill kernel is 1
The support of kernel is 1
Not separable

Wyvill kernel (red)

01/04/2007 State Key Lab of CAD&CG 109

 Other kernels: Modified Wyvill
Kernels

Modified Wyvill kernel also has similar

shape with Gaussian
hNewWyvill(x) = (1-x2)3.
After scale: hWyvill(0)=1
The integral of Modified Wyvill kernel is 1
The support of kernel is 1
Not separable

Modified Wyvill kernel (blue)

01/04/2007 State Key Lab of CAD&CG 110

 Compare the different kernels

Comparison of Filter Kernels after scale

upper (in black): the Gaussian kernel, e-0.69314715x2
middle (in red): the B-spline kernel, (1.5)hBspline(0.72235x)
lower (in blue): the Wyvill kernel, hWyvill(0.5x)

01/04/2007 State Key Lab of CAD&CG 111

 Computation of Convolution

The Gaussian is separable and spherically

symmetric
Integration by part!
B-spline and Wyvill kernel is not separable
summation of point source terms
the product of integration and distance filters.

01/04/2007 State Key Lab of CAD&CG 112

 Computation of Convolution

Approximations to B-
Spline Convolution
left: summation of point
sources
right: product of filters

Approximations to
Wyvill Convolution
left: summation of point
sources
right: product of filters

01/04/2007 State Key Lab of CAD&CG 113

 Computation of Convolution

Gaussian Convolution
approximation for 3, 7,
and 15 Segments (left) or
Points (right)

01/04/2007 State Key Lab of CAD&CG 114

 Ramification Modeling

A Two-Ramiform with
Constant Radii

A Trifurcated Ramiform

01/04/2007 State Key Lab of CAD&CG 115

 Please Enjoy More Examples of
Convolution Surface

01/04/2007 State Key Lab of CAD&CG 116

 Please Enjoy More Examples of
Convolution Surface

Snowflake

01/04/2007 State Key Lab of CAD&CG 117

 Potted plant

01/04/2007 State Key Lab of CAD&CG 118

 Maple tree

01/04/2007 State Key Lab of CAD&CG 119

 Please Enjoy More Examples of
Convolution Surface

A gourd shape

01/04/2007 State Key Lab of CAD&CG 120

 Please Enjoy More Examples of
Convolution Surface

A dinosaur

01/04/2007 State Key Lab of CAD&CG 121

 Please Enjoy More Examples of
Convolution Surface

An cartoon enforcer

01/04/2007 State Key Lab of CAD&CG 122

 Quadric Surfaces

Basics of Quadric Surfaces

Properties of Quadric Surfaces
Representation of Quadric Surfaces
Applications of Quadric Surfaces

01/04/2007 State Key Lab of CAD&CG 123

 Basics of Quadric Surfaces

A general quadric surface is

Ax2 + By2 + Cz2 + 2Dxy + 2Exz + 2Fyz +2Gx +
2Hy + 2Jz + K = 0
Its matrix representation
{x | F(x) = xTQx = 0}
where

01/04/2007 State Key Lab of CAD&CG 124

 Basics of Quadric Surfaces

There are nine types of general quadrics according

coefficients of this algebraic equation
Ellipsoid
Elliptic cone
Cylinder (elliptic, parabolic, hyperbolic)
Hyperboloid (of 1 sheet, of 2 sheets)
Paraboloid (elliptic, hyperbolic)
Degenerated cases
point, plane, parallel planes,
Invalid shapes
imaginary quadric, intersecting imaginary planes, imaginary
parallel planes
01/04/2007 State Key Lab of CAD&CG 125
 Sketches of General Quadric
Surfaces

01/04/2007 State Key Lab of CAD&CG 126

 Natural Quadrics

Natural Quadrics: subset of general

sphere (special case of ellipsoid)
right circular cone (special case of elliptic
cone)
right circular cylinder (special case of elliptic
cylinder)
Planes

01/04/2007 State Key Lab of CAD&CG 127

 Natural Quadrics

Natural quadrics are by far the most popular set

of quadric: Prismatic Solid
Mechanical CAD domains

01/04/2007 State Key Lab of CAD&CG 128

 Quadric Surface Patches

Pieced together with tangent plane (G1)

continuity to create free-form shapes

Some complex free-form solids modeled with quadric algebraic patches: (a) a
genus five solid with 2,400 quadratic algebraic patches, (b) a bone head, the
entire bone was modeled with 5,696 quadratic algebraic patches, and (c) a
knot modeled with 6,912 quadratic algebraic patches.

01/04/2007 State Key Lab of CAD&CG 129

 Quadric Surface Patches

(a) A vase modeled with 336 quadratic algebraic patches, rendered

with transparency to show some of its trunctets. (b) The same vase in
an environment with multiple light sources and participating media.

01/04/2007 State Key Lab of CAD&CG 130

 Properties of Quadric Surfaces

Let QA (QB) denotes the matrix representation of

quadric surface in frame A(B), M denotes an
affine transformation from frame A to from B

Then the transformation of quadrics is

QB=MTQAM

01/04/2007 State Key Lab of CAD&CG 131

 Properties of Quadric Surfaces

Let Qu denotes upper-left 33 sub-matrix of Q

The rank of both Q and Qu are invariant under affine

transformation
The type of quadric surface is invariant under affine
transformations
The quadric surfaces are invariant under rigid
motions

01/04/2007 State Key Lab of CAD&CG 132

 Characteristic Function

Characteristic functions are

Det(Q-I) Det(Qu-I)
Invariant under rigid motion
Used for classifying the type of quadrics
according to its eignvalues

01/04/2007 State Key Lab of CAD&CG 133

 Characteristic Function (1)

01/04/2007 State Key Lab of CAD&CG 134

 Characteristic Function (2)

01/04/2007 State Key Lab of CAD&CG 135

 Types of Quadric Surfaces
Types are decided by using a decision tree

01/04/2007 State Key Lab of CAD&CG 136

 Ruled Quadric Surfaces

Ruled Quadric Surfaces

A family of straight lines can be found which
lie entirely on the quadric surface
cylinders, cones, hyperbolic paraboloid,
hyperboloid of one sheet

01/04/2007 State Key Lab of CAD&CG 137

 Pencils of two Quadric Surfaces

F1(x,y,z) and F2(x,y,z) are two quadric surfaces,

then the pencil of them is
F1(x,y,z)+F2(x,y,z)
where is real scalar
Every quadric surface in the pencil of F1(x,y,z) and
F2(x,y,z) pass the intersection curves of F1(x,y,z) and
F2(x,y,z)
F1(x,y,z)=0 and F2(x,y,z)=0
Thus
F1(x,y,z)+F2(x,y,z)=0
01/04/2007 State Key Lab of CAD&CG 138
 Pencils of two Quadric Surface

At least one ruled quadric in the pencil of any

two general quadrics
Ruled quadrics are easily parameterized
The quadric surface intersection curves can be
expressed parametrically using the
parameterization scheme for the ruled quadric in
the pencil

01/04/2007 State Key Lab of CAD&CG 139

 Representation of Quadric Surfaces

Two representation techniques:

Algebraic
Geometric

01/04/2007 State Key Lab of CAD&CG 140

 Algebraic Representation of
Quadric Surfaces

Represented by ten coefficients of quadric

implicit polynomial equation F(x,y,z):
Advantage: a common set of routines can be
written for handling all types of general
quadric surfaces
Disadvantage: lack of computational
robustness
Floating point data of imperfect accuracy
Internal inconsistency

01/04/2007 State Key Lab of CAD&CG 141

 Geometric Representation of
Quadric Surfaces

Represented by
{1 point, 2 orthogonal unit vectors, 3 scalars}
The point fixes the position of the surface
The vectors define its orientation or axes
The scalars determine its dimensions
Example: Ellipsoid
Its center (a point)
Two of its three orthogonal axes (two orthogonal unit
vectors)
Three lengths (radii) along its three axes (three scalars)

01/04/2007 State Key Lab of CAD&CG 142

 Geometric Representation of
Quadric Surfaces

Advantages: its robustness and internal

consistency
Disadvantage: a large number of routines
are needed to handle all the special cases
that arise as a result of each type of
surface being treated as a separate entity
(e.g. a problem of combinatorial explosion
in the number of intersection routines)

01/04/2007 State Key Lab of CAD&CG 143

 Applications of Quadric
Surfaces
A surface/surface or curve/surface intersection:
computationally and topologically tractable as compared
to parametric patches.
For example,
a quadric surface F(x,y,z)=0
a line L:(x(t),y(t),z(t)),
line and surface intersection F(x(t),y(t),z(t))=0 is a uni-variate
quadratic function.

Quadratic algebraic patches with smooth continuity for

modeling complex free-form shapes

01/04/2007 State Key Lab of CAD&CG 144

 Download Course

[Link]

01/04/2007 State Key Lab of CAD&CG 145