The International Archives of the Photogrammetry, Remote Sensing andSpatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1104
Linear Stretch
Equalization Stretch
Ik-Orig (red)
Ik-Shift (green)
Ik-Str (blue)
Ik-Orig (red)
Ik-Str(green)
Ik-Str-Shift (blue)
Ik-Orig (red)
Ik-Shift (green)
Ik-Str (blue)
Ik-Orig (red)
Ik-Str(green)
Ik-Str-Shift (blue)
Figure 2. Comparison of quality difference between individual bands of the four testing images (Ik-Orig, Ik-Shift, Ik-Str, and Ik-Str-
Shift) by overlaying the same band from different images in different colour and checking colour appearance in the overlaid images
(colour indicating difference exists; no colour indicating no difference)
2.2 Classification
To prove that the four testing images have the same image
quality for digital classification, the ISODATA clustering tool
is selected to cluster the four images into the same number of
clusters using exactly the same clustering parameters. The
ISODATA clustering is selected, instead of any supervised
classifiers, to avoid operator’s influence in the classification.
All of the four spectral bands are used in the clustering. The
images are clustered into 16 clusters. And, the maximum
clustering iteration is 20.
To precisely compare the 16 clusters classified from the four
images, the clustering result from Ik-Orig is shown in Figure
3.a, and the results from Ik-Shift, Ik-Str and Ik-Str-Shift are
overlaid with that of Ik-Orig and displayed in Figure 3.b, 3.c