Camera Models and

Perspective Projection
Camera Models

Prof. Schrater
 Cameras
• First photograph due to Niepce
• First on record shown in the book -
1822
• Basic abstraction is the pinhole camera
– lenses required to ensure image is not too
dark
– various other abstractions can be applied
 Pinhole cameras
• Abstract camera • Pinhole cameras
model - box with a work in practice
small hole in it
 Copyright Christopher Tyler

Vasari’s Portrait of
Filippo Brunelleschi

Edgerton’s depiction of the reverse-reflection ‘peepshow’ of the

Baptistry with which Brunelleschi demonstrated perspective in ~1420
Copyright
Christopher Tyler

The ‘Goldman’
Annunciation

by Masolino

(1424)

This complex
architecture is
estimated to have
predated (and may
have inspired)
Masaccio’s ‘Trinity’
by four years.
Copyright
Copyright
Christopher
ChristopherTyler
Tyler

The accurate perspective extends from the arch in the background

to to the coffers of the portico . . .
Copyright
Christopher Tyler

. . . and even details of the Virgin’s throne in the right foreground.

 Copyright
Christopher Tyler

As required in accurate perspective, the 45º obliques project to

distance points at the same height as the central vanishing point.
Viewpoint dependent
Projections
Samuel von Hoogstraten
 What’s wrong?

Ron Davis
Six-Ninths Blue, 1966
Ames Room
 Geometric properties of
projection
• Points go to points
• Lines go to lines
• Planes go to whole image
• Polygons go to polygons
• Degenerate cases
– line through focal point to
point
– plane through focal point
to line
Distant objects are smaller
 Parallel lines meet
Common to draw film plane
in front of the focal point.
Moving the film plane merely
scales the image.
Polyhedra project to polygons
• (because lines
project to lines)
 Can we invert?
Polygons -> Polyhedra?
• Although not unique,
methods for interpreting Junctions
line drawing exist.
• Key idea:
– Classify junctions
– look for consistent sets of
line labels, bounding
polyhedra
– “line labelling” - e.g.
• Boundary lines
• Corners (convex vertices) Interpret drawing in terms
– Practical problem - too hard of a set of planar facets--
to extract lines and Assign lines to facets
junctions to label from real
images Determine facet normals
Curved surfaces are much more
interesting under projection

• Crucial issue:
outline is the set of
points where the
viewing direction is
tangent to the
surface
• This is a projection
of a space curve,
which varies from
view to view of the
surface
 Vanishing points
• each set of parallel lines • Good ways to spot
(=direction) meets at a faked images
different point – scale and
– The vanishing point for perspective don’t
this direction work
• Sets of parallel lines on – vanishing points
the same plane lead to behave badly
collinear vanishing – supermarket tabloids
points. are a great source.
– The line is called the
horizon for that plane
Faked Photos
 One Point

Three Point

Two Point
 The equation of projection
• If you know P=(x,y,z) and f’, what is P’?
– Where does a point in view space end up on
the screen?
 The equation of projection
y P’(xs,ys) P(x,y,z)

y/z = ys /f

f -z
View Plane

• Cartesian –Ignore the third coordinate, and get

coordinates: x y
(x, y, z) " ( f , f )
– We have, by similar z z
triangles, that
(x, y, z) -> (f x/z, f
y/z, -f)
 Homogenous coordinates
• Add an extra coordinate • Basic notion
and use an equivalence – Possible to represent
relation points “at infinity”
• for 2D • Where parallel lines
intersect
– equivalence relation
• Where parallel planes
k*(X,Y,Z) is the same as
intersect
(X,Y,Z)
– Possible to write the
• for 3D action of a
– equivalence relation perspective camera
k*(X,Y,Z,T) is the same as a matrix
as (X,Y,Z,T)
 The camera matrix

• Turn previous " U % "$ 1 0 %" X %

0 0 '$ '
expression into HC’s $ ' $ ' $Y'
– HC’s for 3D point are $ V ' = $0 1 0 0$ '
' Z
$ ' $ 1
(X,Y,Z,T) #W & # 0 0 f 0 '$$ ''
&# T &
– HC’s for point in
image are (U,V,W)
Relation between HC’s and image coordinates
xs =U/W
ys =U/W
 Weak perspective
• Issue
– perspective effects only
large with depth
variation-
– Perspective effects
frequently small over the
scale of individual objects
– collect points into a group
at about the same depth,
then divide each point by
the depth of its group
– Adv: easy
– Disadv: wrong
Orthographic projection
Orthographic via Telecentric
Lens
The projection matrix for
orthographic projection

"X %
" U % " 1 0 0 0 %$ '
$ ' $ '$Y '
$ V ' = $ 0 1 0 0 '$ '
$ ' $ '$ Z '
#W & # 0 0 0 1 &$ '
#T &
The reason for lenses
Pinhole too big -
many directions are
averaged, blurring the
image

Pinhole too small-

diffraction effects blur
the image

Generally, pinhole
cameras are dark, because
a very small set of rays
from a particular point
hits the screen.
 The thin lens

1 1 1 f= R
- = 2(n-1)
z' z f
z’ = z f/(z+f)
Is limited focus bad?
1) Know your current focal length
2) Blur gradient => Depth
Spherical aberration
Lens systems
Vignetting
 Other (possibly annoying)
phenomena
• Chromatic aberration
– Light at different wavelengths follows different paths; hence,
some wavelengths are defocussed
– Machines: coat the lens
– Humans: live with it
• Scattering at the lens surface
– Some light entering the lens system is reflected off each
surface it encounters (Fresnel’s law gives details)
– Machines: coat the lens, interior
– Humans: live with it (various scattering phenomena are
visible in the human eye)
• Geometric phenomena (Barrel distortion, etc.)
Human Eye-thick lens
CCD Cameras

Noise sources:
1) Spatial response/Efficiency
2) Thermal fluctuations
3) Shot Noise
4) Read Noise (amplifier)
5) Quantization
 Camera parameters
• Issue
– camera may not be at the origin, looking down the z-axis
• extrinsic parameters
– one unit in camera coordinates may not be the same as
one unit in world coordinates
• intrinsic parameters - focal length, principal point, aspect
ratio, angle between axes, etc.
"X %
" U % " Transformation %" 1 0 0 0 %" Transformation %$ '
$ ' $ '$ '$ '$Y '
$ V ' = $ representing '$ 0 1 0 0 '$representing '$ '
$ ' $ '$ '$ '$ Z '
#W & # intrinsic parameters &# 0 0 1 0 &#extrinsic parameters &$ '
#T &
 Camera calibration
• Issues: • Error minimization:
– what are intrinsic – Linear least squares
parameters of the • easy problem
camera? numerically
– what is the camera • solution can be rather
matrix? bad
(intrinsic+extrinsic) – Minimize image distance
• General strategy: • more difficult numerical
problem
– view calibration object • solution usually rather
– identify image points good,
– obtain camera matrix by • start with linear least
minimizing error squares
– obtain intrinsic – Numerical scaling is an
parameters from camera issue
matrix