rs_9fsfssdgdgdgdgdgdgdgsdgdgdgdconverted.pdf

dg

1. ES 7-Principles of
REMOTE SENSING
AND GEOGRAPHIC
INFORMATION
SYSTEM
Engr. Erwin C. Torio, Ph.D.

2. Remote Sensing Course
Module-9
“Image Processing – Conversion”

3. Learning Objective of Module 9
“Image Processing –Conversion”
Objective

To learn about various image conversion and
manipulation techniques for better analysis and
interpretation by understanding the physics of
colors and features.

4. Outline of Module 9
“Image Processing –Conversion”
Contents:

1.Types of Image Conversion
2.Image Enhancement
3.Color Display
4.Spatial filtering
5.Normalized Difference vegetation Index (NDVI)
6.Principal Component Analysis
7.Textural Analysis

5. 1. Types of image conversion

6. Types of image conversion
Image
Conversion
Image
enhancement Color composition
HSI conversion
Grey scale conversion
Histogram conversion
Spectral feature
Principal Component etc
Geometric feature
Edge, Lineament etc.
Textural feature
Spatial Frequency etc.
Feature
extraction
© Shunji Murai 2004

7. Different types of feature in imagery
Spectral pattern
Geometric pattern
Textural pattern



6
Geometric pattern Geometric pattern
Odaiba area
in Tokyo
(reclaimed land)
IKONOS Pan
imagery (1m)

8. Different types of feature in imagery
Fine texture of sea water, grass; Coarse texture of trees
IKONOS Image © JSI, 2003
Textural patterm Textural patterm

9. Different types of feature in imagery
Clear water
Spectral patterm
Road
vegetation
Turbid
water

10. 2. Image Enhancement

11. Grey scale conversions
x
min
y
max
y
min x
y
x
y
x
y
x
y
linear fold
x
max
saw exponential
© Shunji Murai 2004

12. Example:
Grey scale conversion
Example: grey scale conversion on ASTER NIR band
after linear conversion original image

13. Histogram conversions
x
frequency

x
frequency

x
y
© Shunji Murai 2004
x
Histogram Equalization
Histogram Modified Histogram
cumulative Histogram Linear cumulative Histogram
y

14. Example:
Histogram equalization
R : G : B = NIR : RED : GREEN
Before After

15. 3. Color Display

16. Human visions for color
Color matching function
Of CIE1931RGB
Stimulus
value

-0.1
400 500 600
wavelength (nm)
700
b(  )
g(  )
r(  )
435.8 nm 546.1 nm 700 nm
0.4
0.3
0.2
0.1
0
© Shunji Murai 2004

17. Human visions for color
Color matching function
of CIE1931XYZ
Stimulus
value

0.5
0
400 500 600
wavelength (nm)
700
1.5
1.0
2.0
y(  )
z(  )
x(  )
x(  )
© Shunji Murai 2004

18. Color Mixing:
XYZ Color System
780
 X  
Y  
380
780

380
780

380
xLd
yLd
zLd Z  
= constant
L() : spectral irradiance
of standard illumination
() : spectral reflectance
of sample

19. Color Mixing:
XYZ Color System
In 1931,
the CIE ( Commission Internationale de l'Eclairage)
developed the
XYZ color system
Tri-chromatic co-ordinates
X
x 
X  Y  Z
Y
X  Y  Z
y 
Y corresponds brightness
(x,y) corresponds hue and
satuartion

20. 19
Color Representation:
Munsell Color System
white
black

21. Color Space Conversion:
RGB to HSI
Colors in HSI can be derived with respect to
normalized RGB values as ……
The above equations need corrections with the following…
·H = (360
o
- H) if (B/I) > (G/I) and H is normalized by H = H/360
o
·H is not defined if S = 0
·S is undefined if I = 0

22. Color Space Conversion:
RGB to HSI
C
B M
R
I
G Y
White
H
Black
S
Relationship
between RGB
space and
HSI space
© Shunji Murai 2004

23. Color Composite
Mixing two colors
additively leads
to a lighter color
Mixing two colors
substractively leads
to a darker color
G B
Y C
M
Additive Composite
R
Subtractive Composite

24. Natural Color and
Infrared Color Composite
Natural Color Red : Green : Blue = Red : Green : Blue
water
forest
bare
land
city
Landsat ETM+ Imagery

25. Natural Color and
Infrared Color Composite
Infrared Color Red : Green : Blue = IR : Red : Green
water
forest
bare
land
city
Landsat ETM+ Imagery

26. Pseudo Color Display
Different colors may be assigned to the subdivided
gray scale of a single image. Such a color allocation
is called pseudo-color
Greyscale
Rainbow Blue-Green-Red-Yellow
EOS-A

27. Example: Pseudo Color
Grey Scale
Landsat ETM+ Band3

28. Example:
Pseudo Color
Pseudo color Blue-Green-Red-Yellow
Landsat ETM+ Band3

29. Photoshop for Color Display
Red (NIR)
Green (Red)
Blue (Green)
RGB image

30. 4. Spatial filtering

31. Major spatial filters
Filter 3 x 3 operator effects


Sobel
A  B
 1
A 

 2


 1

or
0
0
0
A
2
 B
2
where,
1 1  2
2

B 

0 0
 
1
 
1 2
 

 1 
0


1



gradient
(finite
differences)


Prewitt
A  B
 1
A 

 1


 1

or
0
0
0
A
2
 B
2
where,
1 1 1
1

B 

0 0
 
1
 
1 1
 

1
0


1


gradient
(finite
differences)

32. Major spatial filters
Filter 3 x 3 operator effects
 0

 1


0

 1 0  1
 1

or

 1
 
0
 
 1
 
 1  1
 1


 1


Laplacian 4 8 differential
 1  1
 1
1 
1
9


1

1 1  0
1
 1 
1
 or
5

1
 
0
 
1 0
1


0


Smoothing 1 1
1 1
Median median of 3 x 3 image smoothing

33. Major spatial filters
Filter 3 x 3 operator effects


High-pass
 0

 1


0

 1 0
5  1

 or
 1 0


  1
1 
1
9


1


1
8
1
1
1


1



edge-
enhancement


Sharpening
 1  8
1 
 8 37
9


1  8

1
 8


1



clearer
image

34. Examples of Enhanced Image
ASTER
imagery Median Laplacian
after applying 3 x 3 filter
Sharpen Sobel
R G B = NIR : Red : Green

35. 5. Normalized Difference
Vegetation Index (NDVI)

36. NDVI
NDVI can be defined as,
NIR  R
NIR  R
NIR : near infrared band
R : red band
Image Credit: Earth Observatory, NASA
http://earthobservatory.nasa.gov/

37. 36
Example of NDVI
World NDVI for 2001 from Terra-MODIS
Animation made using data from “Introductory MODIS multi-disciplinary data-set” by NASA

38. 6. Principal Component Analysis

39. Principal Component Analysis
It is used to reduce the dimensions of measured variables
( p dimension ) to the representative principal components
( m dimension )
If p dimensional variables be {x
i } i = 1,p
the principal components {z
k } k = 1, m
can be expressed as the linear combination as follows…
z
k = a
1k x
1 + a
2k x
2 + ...... + a
pk x
p
(a
1k - a
pk ) are determined under the following constrains.
•a
ik = 1
•variance z
k should be maximum
•z
k and z
k+1 should be independent each other
© Shunji Murai 2004

40. Example of PCA for
multi-channel images
x
2
z
x
1
Graphically speaking, the first principal component
for example in the case of two dimensional variables
will be the principal axis which gives the maximum variance.
© Shunji Murai 2004

41. Example of PCA for
multi-channel images
PCA on first 6 bands of Landsat ETM+ Imagery
R G B = NIR : RED : Green

42. Example of PCA for
multi-channel images
PCA on first 6 bands of Landsat ETM+ Imagery
PC 3 PC 1
corresponds
to brightness
PC 2
corresponds
to greenness

43. Example of PCA for
multi-channel images
PCA on first 6 bands of Landsat ETM+ Imagery


R G B
PC 1 2 3

44. 7. Texture Analysis

45. Density and Pattern
Densely spaced
paddy fields
Sparsely spaced
paddy fields
ASTER imagery R G B = NIR : Red : Green

46. Examples of classification
with textural analysis
After textural classification
for bush density
Airphoto of bushes
Red -> Orange -> yellow -> Green -> Blue
Low    High