Image Preprocessing and Information Extraction

1
Image Preprocessing and Information Extraction
Ruiliang Pu Ecosystem Science
Division Department of ESPM University of
California, Berkeley, USA February 24, 2004
2
Outline

• Atmospheric correction
• Geometric correction
• Topographic correction
• Image enhancement
• Information extraction
• Summary

3
Atmospheric correction
4
Atmospheric correction

• A diagram showing atmosphere how to affect on RS
  data

5
Interaction with the Atmosphere
Atmospheric correction
Scattering
Absorption
Scattering occurs when particles or large gas
molecules present in the atmosphere interact with
and cause the electromagnetic radiation to be
redirected from its original path. Absorption is
the other main mechanism at work when
electromagnetic radiation interacts with the
atmosphere. In contrast to scattering, this
phenomenon causes molecules in the atmosphere to
absorb energy at various wavelengths.
6
Mie scattering
Atmospheric correction
7
Rayleigh scattering
Atmospheric correction
Rayleigh scattering occurs when particles are
very small compared to the wavelength of the
radiation.
8
Mie scattering
Atmospheric correction
Mie scattering occurs when the particles are just
about the same size as the wavelength of the
radiation
9
Nonselective scattering
Atmospheric correction
Nonselective scattering. This occurs when the
particles are much larger than the wavelength of
the radiation
10
Atmospheric window It refers to the relatively
transparent wavelength regions of the atmosphere
Atmospheric correction
Those areas of the spectrum which are not
severely influenced by atmospheric absorption
and thus, are useful to remote sensors, are
called atmospheric windows.
11
Atmospheric correction

• Atmospheric correction of ALI imagery (Beijing,
  China, May 3, 2001)

After
Before
After
Before
SL Liang, University of Maryland, College Park
12
Atmospheric correction

• MODIS imagery of Chinese northeastern coast,
  (False color composite, May 7, 2000)

(Before AC)
(After AC)
SL Liang, University of Maryland, College Park
13
Atmospheric correction

• MODIS imagery of Chinese northeastern coast,
  (Natural color composite, May 7, 2000)

(Before AC)
(After AC)
14
Geometric correction
15
Satellite Characteristics Orbits and Swaths
Geometric correction

• Geostationary
• Sun-synchronous

Geostationary Satellites at very high altitudes
(e.g., 36,000 km), which view the same portion of
the Earth's surface at all times have
geostationary orbits. Sun-synchronous Satellite
orbits cover each area of the world at a constant
local time of day called local sun time. At any
given latitude, the position of the sun in the
sky as the satellite passes overhead will be the
same within the same season. Orbit The path
followed by a satellite.
16
Swath
Geometric correction
Swath As a satellite revolves around the Earth,
the sensor "sees" a certain portion of the
Earth's surface. The area imaged on the surface,
is referred to as the swath
17
Geometric distortion sources
Geometric correction

• Sensing principles
• Central Perspective
• Cross-track
• Along-track

18
Geometric correction
Orthographic Central Perspective
19
Geometric distortion sources
Geometric correction

• Platform Status
• Airborne altitude variation, velocity variation,
  attitude (pitch, roll, yaw)
• Satellite earth rotation, earth curvature

20
Geometric correction

• Platform Attitude

http//www.cnr.berkeley.edu/gong/textbook/
21
Geometric distortion sources
Geometric correction

• Map distortion
• deformation of map sheets
• map projection differences
• map scale variation
• map generalization

22
Geometric correction

• Image georeferencing

1) Geometric rectification and image
rectification recovers the imaging geometry 2)
Image-to-image registration refers to
transforming one image coordinate system into
another image coordinating system 3)
Image-to-map registration refers to
transformation of one image coordinate system to
a map coordinate system resulted from a
particular map projection. Georeferencing
generally covers 1) and 3).
23
Georeferencing (geometric correction)
Geometric correction

• A process of transforming from one coordinate
  system (X, Y) (e.g., image), which might be
  distorted due to various factors, to another
  coordinate system (U, V) (e.g., map) with its
  underlying map projection.

24
Geometric correction
Geometric Transformation more general approach

• A first order transformation from image
  coordinate system (x, y) to ground (map)
  coordinate system (U, V, Z)
• In this case, Z is not considered. To estimate
  series of A and B, we need to find at least 3
  linearly independent points with known (x, y) and
  (U, V) coordinates, such points are called ground
  control points (GCPs).
• An explicit geometric transformation would
  require ground points elevation Z (derived
  above) to be known

25
Geometric correction
Geometric Transformation
If the number of GCPs is, i.e., n3, we get a
full rank transformation matrix
UMA ? AM-1U VMB ? BM-1V
26
Geometric Transformation
Geometric correction

• If ngt3
• Using least squares method to solve the
  overdetermined equations
• UMA ? A(MTM)-1MTU

27
Geometric correction
Image Resampling

• Nearest Neighbor
• Bilinear
• Cubic

28
Georeferencing (geometric correction)
Geometric correction
After
Before
29
Geometric correction
Image to Image Registration

• We can also do Image-to-image registration, it
  refers to transforming one image coordinate
  system into another image coordinating system
• We will introduce it in lab

30
Topographic correction
Topographic correction
31
Topographic correction
32
Topographic correction
Aim

• An ideal slope-aspect correction removes all
  topographically induced illumination variation so
  that two objects having the same reflectance
  properties show the same DN despite their
  different orientation to the suns position.

33
Correction Methods
Topographic correction

• 1. Statistic-empirical method, used for
  correcting single scene image

(1)
After
Before
34
Correction Methods
Topographic correction

• 2. Cosine correction, used for correcting
  multitemporal scenes
• Applied in flat terrain to equalize illumination
  differences due to different sun positions in
  multitemporal scenes

sz suns zenith angle
(2)
Note The cosine correction only models the
direct part of the irradiance
Before
After
35
Correction Methods
Topographic correction

• 3. Minnaert correction, used for correcting
  multitemporal or multisensor dataset. Named from
  the Belgian astrophysicist Marell G. J. Minnaert
  (1941)

k Minnaert constant 0, 1
(3)
Note When cos(i) near 0, k increases the
denominator and prevents a division by small
values. Thus one can counteract an
overcorrection in Eq (2)
Eq. (3)
Eq. (2)
36
Correction Methods
Topographic correction

• 4. C-correction, used for correcting
  multitemporal dataset.

(4)
c correction parameter
Note Mathematically, the effect of c is similar
to that of Minnaert constant (k). It increases
the denominator and weakens the overcorrection of
faintly illuminated pixels.
Before
After
37
Image enhancement
38
Histogram
Image enhancement
Image plane 2 Histogram
Image plane 1 Histogram
Image plane 3 Histogram
39
Image stretch and compression
Image enhancement
40
More transform functions (to single band image)
Image enhancement
Higher
Lower
Medium
Exponential
Logarithm
41
(No Transcript)
42
Histogram equalization
Image enhancement
Original composite (NIR,R,G)
Linear stretch
43
Spatial filtering
Image enhancement
LP

• Spatial frequency
• Filtering procedure
• Low-pass filter
• High-pass filter

HP
Spatial filtering encompasses another set of
digital processing functions which are used to
enhance the appearance of an image. Spatial
frequency refers to the frequency of the
variations in tone that appear in an
image. filtering procedure involves moving a
'window' of a few pixels in dimension over each
pixel in the image, applying a mathematical
calculation using the pixel values under that
window, and replacing the central pixel with the
new value
44
Information extraction
Information extraction
45
Through elements of visual interpretation
Information extraction
46
Through image transformations
Information extraction

• Image subtraction
• Spectral ratios
• Vegetation index (VI)
• Normalized difference vegetation index (NDVI)
• NDVI.N
• Linear transformations
• PCA (principal component analysis)
• K-T (Kauth-Thomas) Tasseled cap transform

47
Information extraction
Vegetation indices
 
48
Information extraction
Red
NIR
 
NDVI
49
Information extraction
Principal component analysis
 
50
Information extraction
PCA transformation

Original bands 1 -20
 
PC1
PC2
PC3
51
Information extraction
K-T transform (Tasseled cap transform)
 
1. Redness 2. Greenness 3. Yellowness
52
Information extraction
 
Regression between crown closure () and redness,
greenness, and (G-R)/(GR) derived from K-T
transform
53
Through image classification
Information extraction

• Unsurpervised classification
• Supervised classification

54
Summary

• Atmospheric correction, scattering, absorption,
  atmospheric windows
• Geometric correction, geometric distortion
  sources, Geometric transform, GCP
• Terrain correction, slope-aspect induced radiance
  distortion, four correction methods
• Image enhancement, histogram, stretch,
  compression to single band image, spatial
  frequency, filtering
• Information extraction, visual interpretation,
  image transformations (VIs, PCA, K-T transform,
  etc.)

55
References

• Meyer, P., K. I. Itten, T. Kellenberger, S.
  Sandmeier and R. Sandmeier, 1993, Radiometric
  corrections of topographically induced effects on
  Landsat TM data in an alpine environment, ISPRS
  Journal of P. R. S., 48(4) 17-28.
• Allen, T. R., 2000, Topographic normalization of
  Landsat thematic mapper data in three mountain
  environments, Geocarto International, 15(2)
  13-19.
• Adler-Golden, S. M., M. W. Matthew, G. P.
  Anderson, G. W. Felde, and J. A. Gardner, 2002,
  An algorithm for de-shadowing spectral imagery,
  2002 AVIRIS Workshop Proceedings.

56
References