Conics and Quadrics

9 Appendix
9.5 Conics and Quadrics

In the lecture we will use conics and quadric as geometrical entities. These concepts are
usually not so well known and also not always very intuitive. We experienced that this
is especially confusing at the beginning because the the introduction to projective space is
done at the same time. So it is sometimes better for the beginning, not to try to interpret
these entities but only to know (and accept) the mathematical definitions and how they
are dealt with.
The literature use in this section is Hartley and Zisserman [3] (p. 30, p. 73) and Meyberg
and Vachenauer [4] (Chapter 6, Section 7).
In order to define conics and quadrics we will first introduce the so called quadratic
forms. The conic and quadric are the special cases of quadratic forms for the three and four
dimensional case respectively.
Definition 9.27 (Quadratic Form) We can express any quadratic function f by a square matrix
A
f (x) = xT Ax
It can be shown that we can assume that A is symmetric without loss of generality.
Given a symmetric matrix A then we denote the expression f (x) = xT Ax as Quadratic Form
The fact that the matrix A in the quadratic form f (x) = xT Ax is symmetric has some
nice properties. Given a symmetric matrix A ∈ Rn×n , we have:
• All eigenvalues λi of A are real numbers (λi ∈ R).
• Eigenvectors belonging to different eigenvectors of A are orthogonal.
• algebraic and geometric multiplicities are the same.
This makes it possible for us to transform any conic or quadric to a diagonal matrix. This
is necessary for the classification of the conics and quadrics later on.
9.5.1 Conics
Definition 9.28 (Conic) A Conic is a curve described by a quadratic equation in a plane. For
homogeneous coordinates this is a set of all points x ∈ R3 which satisfy the equation
0 = xT Cx
where C is a symmetric 3 × 3 matrix.
The matrix C is enough to describe a certain conic. Explicitly, it is

x a b/2 d/2
For x =
y
and C =
b/2 c e/2
z d/2 e/2 f
Using this we can write the equation for the conic also as
xT Cx = ax2 + bxy + cy 2 + dxz + exz + f z 2
52 campar - wwwnavab.cs.tum.edu - 3D Computer Vision Script Draft v.000

9.5 Conics and Quadrics
We can see that the conic is a special case of the quadratic form. Thus all properties of
quadratic forms hold for conics as well.
The conic defining matrix C has obviously 6 independent parameters. Since we are
using the projective geometry framework however, all entities are up to scale. Hence a
conic has only 5 degrees of freedom. That is, 5 points define a conic.
9.5.1.1 Intuition and Classification - Euclidean vs. Projective Framework

The geometric meaning of the conic is following. A conic is a quadratic curve in a plain
that we get by intersecting a cone and a plane. Depending on the way we intersect the
plane with the cone we get four different types of the conic. The curves we get can be open
or closed. Check
copyright of
Wiki-Images
(these here are
GNUs)!
Figure 9.1: Conic classification in the Euclidean Framework.Images: http://en.wikipedia.org/wiki/Conic
If the plane is parallel to the axis of the cone, we get a circle, which is a closed curve. If
the plane is not parallel any more but the curve we get by the intersection is still closed,
then the curve is an ellipse. When the plane is parallel to the line which defines the cone
(i.e. to the side of the cone) we get an open curve named parabola. And when the angle gets
even bigger and the plane is not parallel to the side of the cone any more and the curve is
open, it is named hyperbola.
These are the so called nondegenerate conics. Then there are also the so called degenerate
conics which are intersections of the cone and the plane going through the tip of the cone -
resulting in either a point or two lines.
Figure 9.2: Four conic examples
This classical classification however is only valid for the Euclidean geometry framework. It
campar - wwwnavab.cs.tum.edu - 3D Computer Vision Script Draft v.000 53

9 Appendix
can be shown however, that the all nondegenerate four conics types are equivalent under
projective transformations.
It is extremely important to always keep in mind which framework, the Euclidean or
the projective you are using. Especially at the beginning of the Computer Vision lecture,
this is something that is really difficult to get used to.
Using the diagonalization property of the quadric forms we can classify the conics for
the projective framework by the form of the diagonals corresponding to them.
Diagonal of C Conic Type

(1,1,1) This is an improper conic. It has no real points solving the equation.
(1,1,-1) This is a circle. In projective geometry it is equivalent to the circle,
ellipse, parabola and hyperbola form the Euclidean geometry frame-
work.
(1,1,0) This conic is one single point.
(1,-1,0) This gives us the conic consisting of two lines.
(1,0,0) One single line that is counted twice.
9.5.2 Quadrics
Definition 9.29 (Quadric) A Quadric is a set of all points x ∈ R4 which satisfy the equation:
0 = xT Qx
where Q is a symmetric 4 × 4 matrix. So again, the quadric is a special case of the quadratic
forms.
Figure 9.3: Three quadric examples
54 campar - wwwnavab.cs.tum.edu - 3D Computer Vision Script Draft v.000

Conics and Quadrics

Uploaded by

Copyright:

Available Formats

Conics and Quadrics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Conics and Quadrics

Uploaded by

Copyright:

Available Formats

9 Appendix

9.5 Conics and Quadrics

xT Cx = ax2 + bxy + cy 2 + dxz + exz + f z 2

52 campar - wwwnavab.cs.tum.edu - 3D Computer Vision Script Draft v.000

9.5.1.1 Intuition and Classification - Euclidean vs. Projective Framework

Figure 9.1: Conic classification in the Euclidean Framework.Images: http://en.wikipedia.org/wiki/Conic

Figure 9.2: Four conic examples

campar - wwwnavab.cs.tum.edu - 3D Computer Vision Script Draft v.000 53

Diagonal of C Conic Type

Figure 9.3: Three quadric examples

54 campar - wwwnavab.cs.tum.edu - 3D Computer Vision Script Draft v.000

You might also like