750
October 2016
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
Q
uality
E
valuation
SNR, spectral fidelity (or
color quality), and spatial
resolution are the three
most important criteria
used to evaluate the quality
of an image. In this study,
both qualitative and quan-
titative analyses were con-
ducted to compare the above
quality measures between
the pansharpened MS im-
age and the original high
resolution MS image. Four
different
pansharpened
MS images were compared
with their corresponding
high resolution MS images
taken under four different
lighting and exposure con-
ditions (Table 1(c)).
Q
ualitative
C
omparison
The qualitative compari-
son was conducted by dis-
playing the pansharpened
MS image and its corre-
sponding high resolution
MS image using the same
software tool (PCI Geo-
matica) under the same
visualization
conditions,
and then comparing them
side-by-side through visual
inspection of the SNR, color
quality, and spatial details
of the images.
Figures 4 and 5 show two
sections of the original high
resolution MS, 4×4 binned
corresponding low reso-
lution MS, pansharpened
high resolution MS, and
original high resolution
Pan image. The original
high resolution MS image
and Pan image were collect-
ed under 18 lux illuminance
and 100 millisecond (ms)
exposure time. From both
Figure 5. Quality (SNR, color, and detail) comparison between the original HR MS images and the
pansharpened HR MS images for both natural color ([a] vs. [c]) and color infrared ([e] vs. [g]). Corre-
sponding LR MS images are (b) and (f) and the original HR Pan images are (d) and (h) (18 lux, 100 ms).
“the color of the original high resolution MS image appears slightly
different than those of the 4×4 binned low resolution MS image and the
pansharpened MS image, because of the presence of the photon noise in
the original high resolution MS image”
