Page 277 - Digital Analysis of Remotely Sensed Imagery

Image Enhancement 239

(e) (f)

(g)
FIGURE 6.23 (Continued)

It contains only six bands with band six excluded because its spatial
resolution (120 m) is incompatible with the other bands (30 m). The
PCA results of the image are presented in Tables 6.3 to 6.5.
Shown in Table 6.3 is the variance-covariance matrix, which is
symmetric and square. Its dimension equals the number of spectral
bands used in the PCA. The most revealing figures in the matrix are
the main diagonal values. They illustrate the variance of spectral
bands. This variance is indicative of the diversity of pixel values that
are related indirectly to ground covers in the scene. For a given scene,
this variance is also indicative of the potential range of DNs. A band
of a small variance suggests that all pixels have a similar value. There-
fore, it is difficult to separate them spectrally. In this sense, variance
can serve as an indicator of information content of a given spectral
band. Judged against this criterion, the most informative band is