Capturing Light… in man and machine
PHOTOGRAPHY
Image Formation
Sensor Array
Sampling and Quantization
Interlace vs. progressive scan
Progressive scan
Interlace
The Eye
The Retina
Retina up-close
Rod / Cone sensitivity
Electromagnetic Spectrum
More Spectra
Camera White Balancing
Color Sensing in Camera (RGB)
Practical Color Sensing: Bayer Grid
RGB color space
HSV
Programming Project #1
Programming Project #1
3.75M
Категория: ИнформатикаИнформатика

Capturing Light… in man and machine

1. Capturing Light… in man and machine

15-463: Computational Photography
Alexei Efros, CMU, Fall 2012

2. PHOTOGRAPHY

Etymology
PHOTOGRAPHY
light
drawing
/ writing

3. Image Formation

Digital Camera
Film
The Eye

4. Sensor Array

CMOS sensor

5. Sampling and Quantization

6. Interlace vs. progressive scan

http://www.axis.com/products/video/camera/progressive_scan.htm
Slide by Steve Seitz

7. Progressive scan

http://www.axis.com/products/video/camera/progressive_scan.htm
Slide by Steve Seitz

8. Interlace

http://www.axis.com/products/video/camera/progressive_scan.htm
Slide by Steve Seitz

9. The Eye

The human eye is a camera!
• Iris - colored annulus with radial muscles
• Pupil - the hole (aperture) whose size is controlled by the iris
• What’s the “film”?
– photoreceptor cells (rods and cones) in the retina
Slide by Steve Seitz

10. The Retina

Cross­section of eye
Ganglion axons
Ganglion cell layer
Bipolar cell layer
Receptor layer
Cross section of retina
Pigmented
epithelium

11. Retina up-close

Light

12.

Two types of light­sensitive receptors
Cones
   cone­shaped 
   less sensitive
   operate in high light
   color vision
Rods 
   rod­shaped
   highly sensitive
   operate at night
   gray­scale vision
cone
rod
© Stephen E. Palmer, 2002

13. Rod / Cone sensitivity

The famous sock-matching problem…

14.

Distribution of Rods and Cones
# Receptors/mm2
.
lin
d
Fovea B
Spot
150,000 Rods
Rods
100,000
50,000 Cones
Cones
080 60 40 20 0 20406080
Visual Angle (degrees from fovea)
Night Sky: why are there more stars off-center?
© Stephen E. Palmer, 2002

15.

Foundations of Vision, by Brian Wandell, Sinauer Assoc., 1995

16. Electromagnetic Spectrum

Human Luminance Sensitivity Function
http://www.yorku.ca/eye/photopik.htm

17.

Visible Light
Why do we see light of these wavelengths?
10000  C
.
Energy
…because that’s where the
Sun radiates EM energy
5000  C
2000  C
700  C
0
400 700 1000
Visible
Region
2000
Wavelength (nm)
3000
© Stephen E. Palmer, 2002

18.

The Physics of Light
Any patch of light can be completely described
physically by its spectrum: the number of photons 
(per time unit) at each wavelength 400 ­ 700 nm.
# Photons
(per ms.)
400   500    600    700
Wavelength (nm.)
© Stephen E. Palmer, 2002

19.

The Physics of Light
Some examples of the spectra of light sources
# Photons
B. Gallium Phosphide Crystal
# Photons
A.  Ruby Laser
.
400   500    600    700
400   500    600    700
Wavelength (nm.)
Wavelength (nm.)
D.  Normal Daylight
# Photons
# Photons
C.  Tungsten Lightbulb
400   500    600    700
400   500    600    700
© Stephen E. Palmer, 2002

20.

The Physics of Light
% Photons Reflected
Some examples of the reflectance spectra of surfaces
Red
Yellow
Blue
Purple
400          700 400          700 400          700 400          700
Wavelength (nm)
© Stephen E. Palmer, 2002

21.

The Psychophysical Correspondence
There is no simple functional description for the perceived
color of all lights under all viewing conditions, but …...
A helpful constraint:
  Consider only physical spectra with normal distributions
mean
area
# Photons
400
500
variance
600
700
Wavelength (nm.)
© Stephen E. Palmer, 2002

22.

The Psychophysical Correspondence
# Photons
Mean
blue
Hue
green yellow
Wavelength
© Stephen E. Palmer, 2002

23.

The Psychophysical Correspondence
# Photons
Variance
Saturation
hi. high
med. medium
low
low
Wavelength
© Stephen E. Palmer, 2002

24.

The Psychophysical Correspondence
Area
Brightness
# Photons
B.  Area         Lightness
bright
dark
Wavelength
© Stephen E. Palmer, 2002

25.

Physiology of Color Vision
Three kinds of cones:
440
RELATIVE ABSORBANCE (%)
.
530 560  nm.
100
S
M
L
50
400        450      500    550    600  650
WAVELENGTH (nm.)
• Why are M and L cones so close?
• Why are there 3?
© Stephen E. Palmer, 2002

26. More Spectra

metamers

27.

Color Constancy
The “photometer metaphor” of color perception:  
Color perception is determined by the spectrum of light 
on each retinal receptor (as measured by a photometer).
© Stephen E. Palmer, 2002

28.

Color Constancy
The “photometer metaphor” of color perception:  
Color perception is determined by the spectrum of light 
on each retinal receptor (as measured by a photometer).
© Stephen E. Palmer, 2002

29.

Color Constancy
The “photometer metaphor” of color perception:  
Color perception is determined by the spectrum of light 
on each retinal receptor (as measured by a photometer).
© Stephen E. Palmer, 2002

30.

Color Constancy
Do we have constancy over 
all global color transformations?
60% blue filter
Complete inversion
© Stephen E. Palmer, 2002

31.

Color Constancy
Color Constancy:  the ability to perceive the
invariant color of a surface despite ecological
Variations in the conditions of observation.
Another of these hard inverse problems:
    Physics of light emission and surface reflection
    underdetermine perception of surface color
© Stephen E. Palmer, 2002

32. Camera White Balancing

• Manual
• Choose color-neutral object in the photos and normalize
• Automatic (AWB)
• Grey World: force average color of scene to grey
• White World: force brightest object to white

33. Color Sensing in Camera (RGB)

3-chip vs. 1-chip: quality vs. cost
Why more green?
Why 3 colors?
http://www.cooldic
http://www.cooldi tionary.com/words/Bayer-filter.wikipedia
Slide by Steve Seitz

34. Practical Color Sensing: Bayer Grid

Estimate RGB
at ‘G’ cels from
neighboring
values
http://www.cooldictionary.com/
words/Bayer-filter.wikipedia
Slide by Steve Seitz

35. RGB color space

RGB cube
Easy for devices
But not perceptual
Where do the grays live?
Where is hue and saturation?
Slide by Steve Seitz

36. HSV

Hue, Saturation, Value (Intensity)
• RGB cube on its vertex
Decouples the three components (a bit)
Use rgb2hsv() and hsv2rgb() in Matlab
Slide by Steve Seitz

37. Programming Project #1

Prokudin-Gorskii’s Color Photography (1907)

38. Programming Project #1

• How to compare R,G,B channels?
• No right answer
• Sum of Squared Differences (SSD):
• Normalized Correlation (NCC):
English     Русский Правила