PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
October 2016
753
C
olor
Q
uality
To evaluate the color quality, the standard color difference
metric Δ
E
of the International Commission on Illumina-
tion (CIE) was used to measure the difference between
the colors in the original MacBeth Color Checker (Figure
4) and those measured in its image (Imatest, 2016b).
Each color patch in the color checker is measured in the
Lab
Color Space, with
a
and
b
for the color-opponent di-
mensions, and
L
for the lightness dimension. The color
difference, Δ
E,
is calculated using
a
and
b
values in the
original color checker and those measured in its image,
ignoring the influence of the
L
value. When Δ
E
equals
to 1, it corresponds to a just noticeable (color) difference
(JND) (Imatest, 2016c).
Figure 9 shows the average color
difference between the original color
checker and its corresponding im-
ages taken under different lighting
and exposure conditions. The color
difference between the original color
checker and the high resolution MS
image varies from 11 to 20, due to
the fact that the color captured by a
sensor is always somewhat different
from that of the real object. Howev-
er, the color difference between the
original high resolution MS and the
pansharpened high resolution MS
images is in the range of 0.1 to 0.5,
and that between the low resolution
MS and the pansharpened MS is 0.1
to 0.4. It is about 2 to 10 times small-
er than the just noticeable color dif-
ference (Δ
E
= 1). This confirms that
no noticeable color change was intro-
duced by the pansharpening.
S
patial
R
esolution
Spatial resolution (i.e., spatial re-
solving power) often cannot be di-
rectly measured by the pixel size of
an image. An up-sampled low res-
olution image may have a smaller
pixel size than its original image,
but its spatial resolution is still
low. On the other hand, a blurred
image has a lower spatial resolu-
tion than the original clear image,
even though their pixel sizes are
Figure 9. Average color difference between the colors in the MacBeth Color Checker (Figure
4) and those in its corresponding images. The color difference between LR MS and PS HR MS is
0.1 to 0.4; smaller than the just noticeable color difference (ΔE = 1).
Figure 10. ISO Standard Chart (12233:2014) used for MTF50 and 10-90% Rise Distance measure-
ments. A total of 60 slanted edges (indicated by the purple rectangles) were measured in
each spectral band.
40 lux × 20 ms 250 lux × 20 ms 18 lux × 100 ms 18 lux × 250 ms
HR MS
11.4
10.9
16
20.4
LR MS
11.4
10.8
16.1
20.1
PS HR MS
11.2
10.7
16.5
20.5
0
5
10
15
20
25
Δ
E
Image Set
Color Difference
the same. Therefore, in the imaging industry, MTF50
(modulation transfer function at 50% contrast or 50%
contrast spatial frequency) and 10-90% Rise Distance
are commonly used to measure the spatial resolution
of an image. The MTF measures the spatial frequen-
cy response (SFR) of an image (in cycle/pixel), whereas
the 10-90% Rise Distance measures the sharpness of an
edge rising from 10% contrast to 90% contrast (in pixel)
(Imatest, 2016d).
A total of 60 slanted edges in the ISO Standard Chart
(12233:2014) (Figure 10) were measured for each spec-
tral band (Imatest, 2016e); each edge for calculating
one MTF50 value and one 10-90% Rise Distance value.
To reduce the random errors, one average MTF50 val-
ue and one average 10-90% Rise Distance value were
calculated for each image by averaging the 60 MTF50
“no noticeable color change was introduced by
the pansharpening”
