Image Formation
Light can change the image and appearances (images from D. Jacobs) What is the relation between pixel brightness and scene radiance? What is the relation between pixel brightness and scene reflectance ?
2003 by Davi Geiger Computer Vision September 2003 L1.1
Camera Obscura
"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". Da Vinci
[Link] (Russell Naughton)
2003 by Davi Geiger Computer Vision
September 2003
L1.2
Used to observe eclipses (eg., Bacon, 1214-1294) By artists (eg., Vermeer).
2003 by Davi Geiger
Computer Vision
September 2003
L1.3
Cameras
Jetty at Margate England, 1898.
[Link] (Jack and Beverly Wilgus)
2003 by Davi Geiger
First photograph due to Niepce First on record shown in the book - 1822
Computer Vision September 2003 L1.4
Pinhole cameras
Abstract camera model - box with a small hole in it Pinhole cameras work in practice
2003 by Davi Geiger
Computer Vision
September 2003
L1.5
Source emits photons
Light
And then some reach the eye/camera.
Photons travel in a straight line
When they hit an object they: bounce off in a new direction or are absorbed (exceptions later).
2003 by Davi Geiger Computer Vision September 2003 L1.6
Irradiance, E
Light power per unit area (watts per square meter) incident on a surface. If surface tilts away from light, same amount of light strikes bigger surface (less irradiance).
light
surface
2003 by Davi Geiger Computer Vision September 2003 L1.7
Radiance, L
Amount of light radiated from a surface into a given solid angle per unit area (watts per square meter per steradian). Note: the area is the foreshortened area, as seen from the direction that the light is being emitted.
light
surface
2003 by Davi Geiger Computer Vision
September 2003
L1.8
Image Formation
A cos = R2
solid angle subtended by a small patch of area A.
A
L - radiance is the amount of light radiated from a surface per solid angle 2 1 (power per unit area per unit solid angle emitted from a surface. W m sr ) E - irradiance is the amount of light falling in a surface (power per unit area incident in a surface. W m 2 )
2003 by Davi Geiger
Computer Vision
September 2003
L1.9
Surface Radiance and Image Irradiance
A
Pinhole Camera Model
n
f
Same solid angle
A cos I cos = ( z / cos ) 2 ( f / cos ) 2
A cos z = I cos f
2003 by Davi Geiger
Computer Vision
September 2003
L1.10
Surface Radiance and Image Irradiance
A
d f
Solid angle subtended by the lens, as seen by the patch A
d 2 cos d = = cos 3 4 ( z / cos ) 2 4 z
2
Power from patch A through the lens 2
cos 4
September 2003 L1.11
2
Thus, we conclude
2
P = L A cos = LA
d 3 cos cos 4z
P A d d E= =L cos 3 cos = L I I 4 z 4 f
2003 by Davi Geiger Computer Vision
Summary
d E = L cos 4 4 f
The irradiance at the image pixel is converted into the brightness of the pixel Image Irradiance is proportional to Scene Radiance Scene distance, z, does not affect/reduce image brightness (the model is too simplified, since in practice it does.) The angle of the scene patch with respect to the view () reduces cos 4 the brightness by the . In practice the effect is even stronger.
2
2003 by Davi Geiger
Computer Vision
September 2003
L1.12
The Bidirectional Reflectance Distribution Function (BRDF)
n n
i
i
( i , i ) ( e , e )
L( e , e ) f ( i , i , e , e ) = E ( i , i )
BRDF - How bright a surface appears when viewed from one direction while light falls on it from another.
Usually f depends only on e i , i , e : true for matte surfaces and specularly reflecting surfaces.
2003 by Davi Geiger Computer Vision September 2003 L1.13
Extended Light Sources and BRDF
= sin i i i
i
A
Light source radiance arriving through solid angle
E ( i , i ) = E ( i , i ) sin i i i
Power arriving at patch A from
Foreshortening
P = A cos i E ( i , i ) = A E ( i , i ) cos i sin i i i
E0
P = =
A
0
thus the irradiance arriving at patch A is
The radiance of a patch A at direction( e , e ) is thus, given by
E ( i , i ) cos i sin i i i
L( e , e ) =
f ( i , i , e , e ) E ( i , i ) cos i sin i i i
Computer Vision September 2003 L1.14
2003 by Davi Geiger
Special Cases of BRDF
1. Lambertian Surfaces (matte)- appears equally bright from all viewing directions and reflects all incident light, absorbing none, i.e. the BRDF is constant and L = E0. . What constant f ?
L( e , e ) = f
E ( i , i ) cos i sin i i i = f E0
p =
since
Thus, the total reflected power from patch A becomes
2
0
L( e , e ) A cos e sin e e e = f E0 A cos e sin e e e =
we finally obtain Foreshortening
Using that L =
p = f E0 and for Lambertian surfaces L = E0 , A
1 f =
2003 by Davi Geiger
Computer Vision
September 2003
L1.15
Special Cases of BRDF
Specular Surfaces (mirrors) reflects all light arriving from the (into) the direction ( i , i .+The BRDF is in this case ) direction i , i ( e and i ) proportional to the product of two impulses, (e i .What is the factor of proportionality ) k ( i?i ) ,
1.
L( e , e ) = E0 =
= k ( e , e ) E ( e , e ) cos e sin e
2
2
k ( i , i ) ( e i ) (e i ) E ( i , i ) cos i sin i i i
1 and for specular surfaces L = E0 k ( i , i ) = cos i sin i ( e i ) (e i ) we finally obtain f ( i , i , e , e ) = cos i sin i
2003 by Davi Geiger Computer Vision September 2003 L1.16
L =
0 0
E ( i , i ) cos i sin i i i L( e , e ) cos e sin e e e
Lambertian Surface Brightness
How bright will a Lambertian surface be when it is illuminated by a point source of radiance E? and by a sky of uniform radiance E? For a point source the irradiance at the surface is E0 = E cos i and the radiance must then be
( i , i ) ( e , e )
1 L( e , e ) = f E0 = E cos i
Familiar cosine or Lamberts law of reflection from matte surfaces (surfaces covered with finely powdered transparent materials such as barium sulfate or magnesium carbonate), and can approximate paper, snow and matte paint. Finally, for a sky of uniform radiance E we obtain
L( e , e ) =
2003 by Davi Geiger
1 E cos i sin i i i = E !
Computer Vision September 2003 L1.17
Special Cases: Lambertian Examples
Lambertian sphere as the light moves. Scene (Oren and Nayar)
2003 by Davi Geiger Computer Vision September 2003 L1.18
(Steve Seitz)
Lambertian+Specular
([Link]
2003 by Davi Geiger Computer Vision
September 2003
L1.19
