Vous êtes sur la page 1sur 47

Image Analysis

Extracting information
from various types of images

A. Dermanis
image types
B & W (panchromatic) Color R

Multispectral Simultaneous images at 3 visible wavelengths (Red – Green – Blue)

Simultaneous images at various wavelengths (including non-visible ultra-violet and infra-red) A. Dermanis
image types
Overlapping images taken from different points and camera orientations

Close range photography for monument documentation


A. Dermanis
image types
Overlapping images taken from different points and camera orientations

Aerial photography for mapping


A. Dermanis
image types
Images taken at constant time intervals

Video - Cinematography

Computer Vision
A. Dermanis
image types
Images taken at parallel planes

Medical Tomography

A. Dermanis
image types
Analog photographs

- There is a gray value for each point on film/photo


Object

- All gray values are possible


from zero (black) to maximum (white)

Printed
photo

Film
(negative) Focal point

A. Dermanis
image types
Digital photographs

Continuous
object
Discrete
photograph

Focal point

A. Dermanis
image types
Digital photographs

Discrete image Stored


divided into square as an n  m matrix
pixels of gray values

Only discrete integer gray values possible


e.g. from 0 (black) to 255 (white) for 8-bit system (28 = 256)
Generally: from 0 (black) to 2n -1 for n-bit system

A. Dermanis
image types
Digital photographs

Only discrete integer gray values possible


e.g. from 0 (black) to 255 (white) for 8-bit system (28 = 256)
Generally: from 0 (black) to 2n -1 for n-bit system

A. Dermanis
image types
Scanners

Pixel by pixel or row by row digital image registration


A. Dermanis
image types
Scanning imagers

Pixel by pixel or row by row digital image registration


A. Dermanis
image types
Scanning imagers

Pixel by pixel digital image registration


A. Dermanis
image types
Scanning imagers

Row by row digital image registration


A. Dermanis
Information in images
Geometric information

P1
F1

If the position of the camera ( of its focal point F ) and its orientation are known
then the position on the photo plate of the image P of a point M
determines that the point lies on a certain line, the line FP

A. Dermanis
Information in images
Geometric information

P1
P2
F1
F2

From the known positions of the images P1 and P2 of an object point M on two photos
with known position and orientation,
the position of the point is determined as the intersection of two lines F1P1 and F2P2

A. Dermanis
Information in images
Geometric information

Photogrammetry
photo – φως = light
P1
gram – γράφω = draw P2
F1
metry – μετρώ = measure F2

From the known positions of the images P1 and P2 of an object point M on two photos
with known position and orientation,
the position of the point is determined as the intersection of two lines F1P1 and F2P2

A. Dermanis
Information in images
Qualitative information

Thematic map

Land cover types


from the classification
of 6 bands of Landsat TM
satellite imagery

A. Dermanis
Information in images
Qualitative information

Thematic map

Land cover types


from the classification
of 6 bands of Landsat TM
satellite imagery

Remote Sensing
A. Dermanis
Information in images
Qualitative information

Thermal mapping

Temperatures from
Infrared imagery

A. Dermanis
Information in images
Human eye-brain vs computer

Human perception: Image as a hole Computer analysis: based on pixel value

A. Dermanis
Information in images
Human eye-brain vs computer

Human perception: Image as a hole Computer analysis: based on pixel value

Recognition of lines by color contrast. Similar processes in computer analysis


Composition of lines into shapes. are a real challenge.
Recognition of objects (window, tree) Necessary for automatic image processing.
by comparison with prototypes in memory

Towards computer vision – artificial inteligence


A. Dermanis
Image Analysis
Image processing
Enhancement of informational content of a single image

Photogrammetry
Extraction of geometric information from multiple images

Remote Sensing
Extraction of qualitative (thematic) information from multispectral images

A. Dermanis
Color
“Color” is the way that the human brain realizes
after visual stimulation
the differences in wavelengths
of a very small part of the electromagnetic radiation.

The “visible” part of the electromagnetic spectrum


(what we usually call light)
corresponds to the central and largest part
of the solar radiation that penetrates the earth atmosphere,
within which the human species has evolved.

A. Dermanis
Electromagnetic radiation

ΔΩ

P
ΔΑ Sensors collect electromagnetic energy ΔQ
emitted from a surface area ΔΑ (pixel),
during a time interval Δt,
arriving at the sensor aperture
with a solid angle ΔΩ

Το characterize the “intensity”


of electromagnetic radiation
we must get rid of ΔΑ, Δt and ΔΩ !
A. Dermanis
Sensors collect electromagnetic energy ΔQ
emitted from a surface area ΔΑ (pixel),
during a time interval Δt,
arriving at the sensor aperture
with a solid angle ΔΩ

Το characterize the “intensity” ΔΩ


of electromagnetic radiation
we must get rid of ΔΑ, Δt and ΔΩ ! P
ΔΑ
Basic definitions (Q = energy)

Q dQ
radiant flux Φ(t):     (power !)
t dt
 d
radiant exitance M(t,P): M   M (emitted)
A dA
 d
irradiance E(t,P): E   E (incident)
A dA
E
illuminance L: L (π = half upper space)

The electromagnetic radiation
Electromagnetic signals x(t)
consist of sines and cosines

sin(ωt ) = sin(2π / T ) = sin(2πc / λ )


cos(ωt ) = cos(2π / T ) = cos(2πc / λ )

with varying periods T, or


angular frequencies ω = 2π/Τ,
or wavelengths λ = cT (c = light velocity)




1
Fourier analysis: x(t )  X ( ) eit dt
2


eiωt = cos(ωt ) + i sin(ωt )


A. Dermanis
 

signal power: P  lim


  

[ x(t )]2 dt 
 S ()d


S (ω) = power spectral density function




1
radiant flux (power):   ( ) d 
2



exitance (with ωλ = cT = 2πc/ω): M 
M

 ( ) d 

dM
M  = spectral exitance
d
A. Dermanis
Sensors respond to exitance only within a spectral band λ1  λ  λ2 :

2

Ideal sensor: M [1 ,2 ] 


M
1
 ( ) d 

2

Actual sensor: M [1 ,2 ] 


M
1
 ( ) w( ) d 

w(λ) = sensor sensitivity response function

response functions
for the 4 sensors
of the Landsat satellite
Multispectral Scanner

A. Dermanis
The electromagnetic spectrum
A cm λ
0.1 1 10 102 103 104 105 106 0.1 1 10 102 103 104 105 106 107

A μ cm m km
0.3 0.2 3 30 300 0.3 3 30 300 0.3 3 30 3 30 300 3 30 300

RADAR
γ
RADIO AUDIO AC
Χ
MICROWAVES

UV IR

VISIBLE

UV (Ultraviolet)  Violet Red  IR (Infrared)

A. Dermanis
Laws of Electromgnetic Radiation

black body : Law of Plank:


an idealized body absorbing (spectral exitance of black body)
all wavelengths of incident radiation
c1
or emitting radiation at all wavelengths M λ ,b ( λ ,T ) = c2
Physical approximation = sun! λ5 e λT
( - 1)
T = temperature
Law of Stefan-Bolzman:
(total spectral exitance)


M b (T )  M  ,b ( , T )d    T 4
0

Law of Wien:
(λ of maximal spectral exitance)
c3
λmax =
T
M λ ,b ( λmax ,T ) = max M λ ,b ( λ ,T )
λ
A. Dermanis
The solar electromagnetic radiation

solar irradiance below atmosphere

atmospheric absorption

A. Dermanis
Color combination
Additive Subtractive
(RGB system, computer screens) (filters, paint mix)

A. Dermanis
Color combination
Additive Subtractive
(RGB system, computer screens) (filters, paint mix)

A. Dermanis
Color combination
Additive Subtractive
(RGB system, computer screens) (filters, paint mix)

Green

Red Blue

A. Dermanis
Color combination
Additive Subtractive
(RGB system, computer screens) (filters, paint mix)

Green

Yellow Cyan
White

Red Magenta Blue

A. Dermanis
Color combination
Additive Subtractive
(RGB system, computer screens) (filters, paint mix)

Green Yellow

Yellow Cyan
White

Red Magenta Blue Magenta Cyan

A. Dermanis
Color combination
Additive Subtractive
(RGB system, computer screens) (filters, paint mix)

Green Yellow

Yellow Cyan Red Green


White Black

Red Magenta Blue Magenta Blue Cyan

CMYK (Cyan Magenta Yellow Black)


A. Dermanis
The RGB Color System

The color Cube

A. Dermanis
The 6 faces of the RGB color cube

B R G

G B R
R = 256 G = 256 B = 256

B R G

G B R
R=0 G=0 B=0
Color combination
The CMYK color system
Cyan - Magenta - Yellow - Black)

Used for printing

Yellow

Red Green

Black

Magenta Blue Cyan

A. Dermanis
Color combination
The CMYK color system
Cyan - Magenta - Yellow - Black)

Used for printing

Use of Cyan - Magenta - Yellow inks

No use of black ink

A. Dermanis
Color combination
The CMYK color system
Cyan - Magenta - Yellow - Black)

Used for printing

Use of Cyan - Magenta – Yellow - Black inks

Usual 4-ink printing

A. Dermanis
Color combination
The HIS color system
Intensity I (% of white) and saturation S (% of color) for various angular values of hue H (color)

(also HBS, HLS, HVS: B = Brightness, L = Lightness, V = Value)

Cyclic arrangement
of hues (colors)

A. Dermanis
Color combination
The HIS color system
Intensity I (% of white) and saturation S (% of color) for various angular values of hue H (color)

(also HBS, HLS, HVS: B = Brightness, L = Lightness, V = Value)

A. Dermanis
The HIS color system
Intensity I (% of white) and saturation S (% of color) for various angular values of hue H (color)
End

Vous aimerez peut-être aussi