CS 4495 Computer Vision A.
Bobick
CS 4495 Computer Vision
Camera Model
Aaron Bobick
School of Interactive Computing
Camera Model
CS 4495 Computer Vision A. Bobick
Camera Model
Administrivia
Problem set 1:
How did it go?
What would have made it better?
New problem set will be out by Thursday or Friday, due
Sunday, Sept 22nd, 11:55pm
Today: Camera models and cameras
FP Chapter 1 and 2.1-2.2
CS 4495 Computer Vision A. Bobick
What is an image?
Up until now: a function a 2D pattern of intensity values
Today: a 2D projection of 3D points
Figure from US Navy Manual of Basic Optics and Optical
Instruments, prepared by Bureau of Naval Personnel. Reprinted
by Dover Publications, Inc., 1969.
Camera Model
Camera Model
CS 4495 Computer Vision A. Bobick
First Known Photograph
View from the Window at le Gras,
Joseph Nicphore Nipce 1826
Reproduction, 1952
Heliograph- a pewter plate coated with bitumen of Judea (an asphalt derivative of
petroleum); after at least a day-long exposure of eight hours, the plate was removed and
the latent image of the view from the window was rendered visible by washing it with a
mixture of oil of lavender and white petroleum which dissolved away the parts of the
bitumen which had not been hardened by light. Harry Ransom Center UT Austin
CS 4495 Computer Vision A. Bobick
Camera Model
What is a camera/imaging system?
Some device that allows the projection of light from 3D
points to some medium that will record the light pattern.
A key to this is projection
CS 4495 Computer Vision A. Bobick
Projection
Camera Model
CS 4495 Computer Vision A. Bobick
Projection
Camera Model
CS 4495 Computer Vision A. Bobick
Image formation
Lets design a camera
Idea 1: put a piece of film in front of an object
Do we get a reasonable image?
Camera Model
CS 4495 Computer Vision A. Bobick
Pinhole camera
Add a barrier to block off most of the rays
This reduces blurring
The opening known as the aperture
How does this transform the image?
Camera Model
CS 4495 Computer Vision A. Bobick
Camera Model
Camera Obscura (Latin: Darkened Room)
The first camera
Known to Aristotle (384-322 BCE)
According to DaVinci When images of illuminated objects ...
penetrate through a small hole into a very dark room ... you will
see [on the opposite wall] these objects in their proper form and
color, reduced in size, in a reversed position, owing to the
intersection of the rays".
Depth of the room is the focal length
How does the aperture size affect the image?
Camera Model
CS 4495 Computer Vision A. Bobick
Home-made pinhole camera
Why so
blurry?
http://www.debevec.org/Pinhole/
CS 4495 Computer Vision A. Bobick
Shrinking the aperture
Why not make the aperture as small as possible?
Less light gets through
Diffraction effects
Camera Model
CS 4495 Computer Vision A. Bobick
Shrinking the aperture
Camera Model
Camera Model
CS 4495 Computer Vision A. Bobick
Adding a lens and concept of focus
circle of
confusion
A lens focuses light onto the film
There is a specific distance at which objects are in focus
other points project to a circle of confusion in the image
Changing the shape of the lens changes this distance
Camera Model
CS 4495 Computer Vision A. Bobick
Lenses
F
focal point
optical center
(Center Of Projection)
A lens focuses parallel rays onto a single focal point
focal point at a distance f beyond the plane of the lens
f is a function of the shape and index of refraction of the lens
Aperture of diameter D restricts the range of rays
aperture may be on either side of the lens
Lenses used to be typically spherical (easier to produce) but now
many aspherical elements designed to improve variety of
aberrations
CS 4495 Computer Vision A. Bobick
Camera Model
Thin lenses
Thin lens equation:
Any object point satisfying this equation is in focus
What is the shape of the focus region?
How can we change the focus region?
Thin lens applet: http://www.phy.ntnu.edu.tw/java/Lens/lens_e.html (by Fu-Kwun Hwang )
Slide by Steve Seitz
CS 4495 Computer Vision A. Bobick
The thin lens
y z '
=
y
z
Computer Vision - A Modern Approach
Set: Cameras
Slides by D.A. Forsyth
Camera Model
Camera Model
CS 4495 Computer Vision A. Bobick
The thin lens
y z ' y z ' f
=
=
f
y
z y
Computer Vision - A Modern Approach
Set: Cameras
Slides by D.A. Forsyth
z ' z ' f
=
z
f
Camera Model
CS 4495 Computer Vision A. Bobick
The thin lens equation
z ' z ' f
1
1 1
1 1 1
=
=
=
z z f
z
f
z f z
Computer Vision - A Modern Approach
Set: Cameras
Slides by D.A. Forsyth
Any object point satisfying this equation is in focus.
CS 4495 Computer Vision A. Bobick
Camera Model
Whats in focus and whats not?
circle of
confusion
A lens focuses light onto the film
There is a specific distance at which objects are in focus
other points project to a circle of confusion in the image
Aside: could actually compute distance from defocus
Changing the shape or relative locations of the lens elements
changes this distance
CS 4495 Computer Vision A. Bobick
Camera Model
Varying Focus
Ren Ng
Camera Model
CS 4495 Computer Vision A. Bobick
Depth of field
f / 5.6
f / 32
Changing the aperture size affects depth of field
A smaller aperture increases the range in which the object is
approximately in focus
But small aperture reduces amount of light need to increase
exposure
Flower images from Wikipedia
http://en.wikipedia.org/wiki/Depth_of_field
CS 4495 Computer Vision A. Bobick
Camera Model
Varying the aperture
Large apeture = small DOF
Small apeture = large DOF
CS 4495 Computer Vision A. Bobick
Nice Depth of Field effect
Camera Model
CS 4495 Computer Vision A. Bobick
Field of View (Zoom)
Camera Model
CS 4495 Computer Vision A. Bobick
Field of View (Zoom)
Camera Model
CS 4495 Computer Vision A. Bobick
FOV depends on Focal Length
Smaller FOV = larger Focal Length
Camera Model
CS 4495 Computer Vision A. Bobick
Camera Model
Zooming and Moving are not the same
Camera Model
CS 4495 Computer Vision A. Bobick
Field of View / Focal Length
Large FOV, small f
Camera close to car
Small FOV, large f
Camera far from the car
CS 4495 Computer Vision A. Bobick
Perspective and Portraits
Camera Model
CS 4495 Computer Vision A. Bobick
Perspective and Portraits
Camera Model
Camera Model
CS 4495 Computer Vision A. Bobick
Dolly Zoom
Move camera while zooming, keeping foreground
stationary
Pioneered by Hitchcock in Vertigo (1958)
Original(YouTube link) (2:07)
Widely used (YouTube link)
An Actual Slide by Jim Rehg
Camera Model
CS 4495 Computer Vision A. Bobick
Effect of focal length on perspective effect
From Zisserman & Hartley
CS 4495 Computer Vision A. Bobick
Camera Model
But reality can be a problem
Lenses are not thin
Lenses are not perfect
Sensing arrays are almost perfect
Photographers are not perfect except some of us
Camera Model
CS 4495 Computer Vision A. Bobick
Geometric Distortion
No distortion
Pin cushion
Barrel
Radial distortion of the image
Caused by imperfect lenses
Deviations are most noticeable for rays that pass through the edge
of the lens
CS 4495 Computer Vision A. Bobick
Camera Model
Modeling geometric distortion
Assume project
to normalized
image coordinates
Apply radial distortion
Apply focal length
translate image center
To model lens distortion
Use above projection operation instead of standard projection
matrix multiplication (which you havent seen yet!)
Camera Model
CS 4495 Computer Vision A. Bobick
Correcting radial distortion
from Helmut Dersch
CS 4495 Computer Vision A. Bobick
Chromatic Aberration
Rays of different wavelength
focus in different planes
Can be significantly improved by
undistorting each channel
separately
Camera Model
CS 4495 Computer Vision A. Bobick
Camera Model
Vignetting
Some light misses the lens or is otherwise blocked by parts
of the lens
Camera Model
CS 4495 Computer Vision A. Bobick
Lens systems
Nikon 24-70mm zoom
Real lenses combat these effects with multiple elements.
Computer modeling has made lenses lighter and better.
Special glass, aspherical elements, etc.
CS 4495 Computer Vision A. Bobick
Retreat to academia!!!
We will assume a pinhole model
No distortion (yet)
No aberrations
Camera Model
CS 4495 Computer Vision A. Bobick
Camera Model
Modeling projection coordinate syste
We will use the pin-hole model
as an approximation
Put the optical center (Center Of
Projection) at the origin
STANDARD (x,y)
COORDINATE SYSTEM
Put the image plane (Projection
Plane) in front of the COP
Why?
The camera looks down the
negative z axis
we need this if we want right-
handed-coordinates
CS 4495 Computer Vision A. Bobick
Camera Model
Modeling projection
Projection equations
Compute intersection with
Perspective Projection of ray
from (x,y,z) to COP
Derived using similar triangles
We get the projection by
throwing out the last
coordinate:
Distant objects
are smaller
CS 4495 Computer Vision A. Bobick
Distant objects appear smaller
Camera Model
Camera Model
CS 4495 Computer Vision A. Bobick
Homogeneous coordinates
Is this a linear transformation?
No division by Z is non-linear
Trick: add one more coordinate:
homogeneous image (2D)
coordinates
homogeneous scene (3D)
coordinates
Converting from homogeneous coordinates
Homogenous coordinates invariant under scale
Camera Model
CS 4495 Computer Vision A. Bobick
Perspective Projection
Projection is a matrix multiply using homogeneous
coordinates:
0
1 0
0 1
0
0 0 1/ f
x
0 x
y
f x, f y
y
=
0
z
z
z
0 z / f
1
( u, v )
This is known as perspective projection
The matrix is the projection matrix
The matrix is only defined up to a scale
f is for focal length used to be d
S. Seitz
Camera Model
CS 4495 Computer Vision A. Bobick
Perspective Projection
How does scaling the projection matrix change the transformation?
0
1 0
0 1
0
0 0 1/ f
f
0
0
f
0
x
0 x
y
x
0 = y f , f
z
z
0 z / f
1
x
0 0
y
0 0
z
1 0
1
fx
= fy
z
x y
f , f
z z
S. Seitz
CS 4495 Computer Vision A. Bobick
Geometric properties of projection
Points go to points
Lines go to lines
Polygons go to polygons
Planes go to planes (or half planes)
Degenerate cases:
line in the world through focal point yields point
plane through focal point yields line
Camera Model
CS 4495 Computer Vision A. Bobick
Camera Model
Parallel lines in the world meet in the image
Vanishing point
Camera Model
CS 4495 Computer Vision A. Bobick
Parallel lines converge in math too
Line in 3-space
x(t ) = x0 + at
y (t ) = y0 + bt
z (t ) = z0 + ct
In the limit as t
we have (for c 0 ):
Perspective projection of the line
fx f (x 0 + at)
=
x'(t) =
z0 + ct
z
fy f (y 0 + bt)
=
y'(t) =
z0 + ct
z
fa
fb
x '(t ) , y '(t )
c
c
This tells us that any set of parallel lines (same a, b,
c parameters) project to the same point (called the
vanishing point). What does it mean if c=0?
Camera Model
CS 4495 Computer Vision A. Bobick
Vanishing points
Each set of parallel lines
(=direction) meets at a
different point
The vanishing point for this
direction
Sets of parallel lines on the
same plane lead to
collinear vanishing points.
Good ways to spot faked
The line is called the horizon
scale and perspective dont
for that plane
images
work
vanishing points behave
badly
supermarket tabloids are a
great source.
Camera Model
CS 4495 Computer Vision A. Bobick
Vanishing points
H VPL
VPR
VP2
VP1
Different directions correspond
to different vanishing points
VP3
Camera Model
CS 4495 Computer Vision A. Bobick
Human vision: Mller-Lyer Illusion
Which line is longer?
http://www.michaelbach.de/ot/sze_muelue/index.html
Camera Model
CS 4495 Computer Vision A. Bobick
Other projection models: Orthographic
projection
( x, y , z ) ( x, y )
Camera Model
CS 4495 Computer Vision A. Bobick
Orthographic projection
Special case of perspective projection
Distance from the COP to the PP is infinite
Image
World
Good approximation for telephoto optics
Also called parallel projection: (x, y, z) (x, y)
Whats the projection matrix?
Camera Model
CS 4495 Computer Vision A. Bobick
Other projection models: Weak perspective
Issue
Perspective effects, but
not 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: only approximate
fx fy
( x, y, z ) ,
z0 z0
1 0 0
0 1 0
0 0 0
0
y
0 =
z
1
s 1
x
y ( sx, sy )
1 s
Camera Model
CS 4495 Computer Vision A. Bobick
Three camera projections
3-d point
(1) Perspective:
(2) Weak perspective:
(3) Orthographic:
2-d image position
fx fy
( x, y , z ) ,
z z
fx fy
( x, y, z ) ,
z0 z0
( x, y , z ) ( x, y )
