PE&RS October 2016 Public - page 754

754
October 2016
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
values and 60 10-90% Rise Distance
values, respectively (Figure 11 and
12).
In Figure 11, the MTF50 values for
all the images are almost the same;
around 0.21 cycle/pixel. This means
that the spatial resolutions of all the
images are almost the same. When
taking a close look at the MTF val-
ues, it can be seen that the maximal
MTF difference between HR MS and
HR Pan is 0.014, slightly larger than
that between HR Pan and PS HR MS
(0.011). Since the spatial resolutions
of the HR MS image and HR Pan im-
age are the same (they were taken by
the same sensor under the same con-
ditions), it can be concluded that the
spatial resolutions of the HR Pan im-
age and PS HR MS image are practi-
cally the same too.
For 10-90%Rise Distance, the smaller
the value, the sharper the edge, thus
a better spatial resolution. In Figure
12, three out of four pansharpened
images have slightly smaller values
than those of the corresponding high
resolution MS images, and only one
has a slightly larger value. This can
also lead to the conclusion that the spatial resolution of
the pansharpened MS images is practically the same as
that of the original high resolution MS images.
C
onclusion
and
D
iscussion
In this study, the qualitative and quantitative evalua-
tions of the images taken in a controlled lab setting and
fused using Fuze Go demonstrated:
The SNR of the high resolution Pan images is
significantly higher than (about twice) that of the
high resolution MS images (Figure 8), because
plenty of photons were blocked by the color filters
of the MS sensor.
The SNR of the pansharpened MS images is
equal to or almost the same as that of the high
resolution Pan images (Figure 8), hence it is sig-
nificantly better than that of the original high
resolution MS images.
The color of the pansharpened MS images has no
noticeable difference from the low resolution MS
images and high resolution MS images (Figure 9).
The spatial resolution of the pansharpened im-
ages is basically the same as that of the high
resolution Pan images and high resolution MS
images (Figures 10 and 11).
These findings demonstrated that the combination of a
low resolution MS sensor, a high resolution Pan sen-
sor, and a high quality pansharpening technique can
produce high quality high resolution pansharpened MS
images, that have the same SNR as the high resolution
Figure 12. Average 10-90% Rise Distance measured from the 60 slanted edges in the ISO Stan-
dard Chart (12233:2014). For each image set, the ratio of the 10-90% Rise Distances between
HR MS and PS HR MS is equal to 1 ± 0.002 to 0.025. A ratio of 1 means that the spatial resolu-
tions of the two images are the same.
40 lux x 20 ms 250 lux x 20 ms 18 lux x 100 ms 18 lux x 250 ms
HR MS
2.612
2.573
2.703
2.732
PS HR MS
2.606
2.638
2.682
2.718
Ratio
1.002302379
0.975360121
1.007829978
1.005150846
0
0.5
1
1.5
2
2.5
3
Pixel
Image Set
10-90% Rise Distance (Pixel)
40 lux × 20 ms
250 lux × 20 ms
18 lux × 100 ms
18 lux × 250 ms
HR MS
0.2133
0.2183
0.2048
0.2012
HR Pan
0.2115
0.2177
0.2158
0.215
PS HR MS
0.2086
0.2139
0.2088
0.2036
0
0.05
0.1
0.15
0.2
0.25
Cycle/Pixel
Image Set
MTF50
Figure 11. Average MTF50 values measured from the 60 slanted edges in the ISO Standard
Chart (12233:2014). The maximal MTF difference between HR MS and HR Pan is 0.014; whereas
that between HR Pan and PS HR MS is 0.011, meaning they have practically the same spatial
resolution.
“the spatial resolution of the
pansharpened MS images is
practically the same as that of
the original high resolution MS
images”
739...,744,745,746,747,748,749,750,751,752,753 755,756,757,758,759,760,761,762,763,764,...822
Powered by FlippingBook