Image Pre-Processing

1
Image Pre-Processing Continuation Spectral
Enhancement
2
ImagePre-Processing
Image Pre-Processing
Consists of processes aimed at the geometric and
radiometric correction, enhancement or
standardization of imagery to improve our ability
to interpret qualitatively and quantitatively
image components.

• Radiometric Enhancement
• Image Restoration
• Atmospheric Correction
• Contrast Enhancement
• Solar Angle Adjustment
• Conv. to Exo-Atmos. Reflectance
• Spectral Enhancement
• Spectral Indices
• PCA, IHS, Color Transforms
• T-Cap, BGW

• Spatial Enhancement
• Focal Analysis
• Edge-Detection
• High/Low Pass Filters
• Resolution Merges
• Statistical Filtering
• Adaptive Filtering
• Texture Filters
• Geometric Correction
• Polynomial Transformation
• Ground Control Points
• Reprojections

3
Principal Component Analysis PCA
Principal Components Analysis is a procedure for
transforming a set of correlated variables into a
new set of uncorrelated variables. This
transformation is a rotation of the original axes
to new orientations that are orthogonal to each
other with little or no correlation between
variables
Where digital image processing is concerned, this
procedure is predominantly exploratory in nature
and is used to help in the extraction of features
and to reduce dimensionality of data
4

• This scatterplot between two spectral bands
  implies a strong correlation. One band can be
  used to predict (to a certain level) the response
  of the other.
• Principal Component Analysis PCA
• Reduces these data into two orthogonal
  components. The first (CI) contains the common
  information between bands 1 and 2. The second
  (CII) contains residual, or independent,
  information.
• Depending on the amount of covariance between
  bands 1 and 2, the second component may not
  contain a significant amount of information and
  can be eliminated.

After Lillesand and Keifer, 1994
Source http//umbc7.umbc.edu/tbenja1/exer1.html
5
Landsat Thematic Mapper image of Curlew Valley
taken on July 4th, 1999 with a 4,3,2 (RGB) band
combination.
6
Eigenmatrix
The Eigenmatrix contains the coefficients used to
calculate each component for the input image.
This matrix is a direct result of the covariance
between each band.
Eigenvalues
The Eigenvalues show the amount of information
contained within each component.
7
Factor loadings what type of component (i.e.
visible, infrared) is it?
Rkp Factor loading akpEigenvector for band k
and component p ?p Eigenvalue for the pth
component Sk Standard deviation for band k
8
(No Transcript)
9
Landsat Thematic Mapper image of Curlew Valley
taken on July 4th, 1999 converted to a principal
component image with a 1,2,3 (RGB) PCA channel
combination.
10
PCA1 (TM-B1)0.194 (TM-B2)0.0413
(TM-B3)-0.332 (TM-B4)0.196 (TM-B5)-0.078
(TM-B6)0.640 (TM-B7)-0.628
Accounts for - 72.28 of variation
PCA2 (TM-B1)0.222 (TM-B2)0.102
(TM-B3)-0.406 (TM-B4)0.220 (TM-B5)-0.142
(TM-B6)0.370 (TM-B7)0.754
Accounts for 22.07 of variation
PCA3 (TM-B1)0.348 (TM-B2)0.092
(TM-B3)-0.495 (TM-B4)0.312 (TM-B5)-0.209
(TM-B6)-0.670 (TM-B7)-0.184
Accounts for 2.62 of variation
11
PCA4 (TM-B1)-0.232 (TM-B2)0.962
(TM-B3)0.045 (TM-B4)0.071 (TM-B5)0.105
(TM-B6)-0.011 (TM-B7)-0.034
Accounts for 1.84 of variation
PCA5 (TM-B1)0.556 (TM-B2)0.202
(TM-B3)0.439 (TM-B4)-0.289 (TM-B5)-0.608
(TM-B6)0.050 (TM-B7)-0.009
Accounts for 0.90 of variation
PCA5 (TM-B1)0.201 (TM-B2)-0.053
(TM-B3)0.533 (TM-B4)0.801 (TM-B5)0.170
(TM-B6)0.017 (TM-B7)0.023
Accounts for 0.22 of variation
12
PCA5 (TM-B1)0.622 (TM-B2)-0.094
(TM-B3)-0.022 (TM-B4)-0.289 (TM-B5)0.720
(TM-B6)-0.011 (TM-B7)0.018
Accounts for 0.07 of variation
Components 1 3 account for 96.97 of the total
variation
While the remaining components only account for a
combined 3.03 of the total variation, there may
be spatial information available.