ISPRS Commission III, Vol.34, Part 3A „Photogrammetric Computer Vision“, Graz, 2002
4. Using scanned aerial imagery with different sharp-
ness, caused by the scanning procedure, we test whe-
ther the method also reacts to natural differences in
sharpness (cf. fig. 8).
In all cases the minimum resolving power of an ideal edge
detector is given. In the case of digital images we refer to
a pixel size of 15 um.
3.1 Demonstration on synthetic Data
Test on noiseless data. The following sequence of grad-
ually blurred images was used to test the proposed method
to determine the point spread function and the resolving
power with respect to correctness of the implemented al-
gorithm.
T T T T
— 1 m
© ; E
e 0.5 - 4 z 5
© 8 N
— ©
0.5 |- $ 9 d m
Il ee] ; S
o its | | 1 1 j d
-1:70.: 70:705 «1
T T T
— 1 [ X
©
a 0s aa ka i 5
0 d
e [y e
I 0.5: > Jor
o > i | i 1 j 8
1 -05 0.05 |
MT T T
— 1 m 7|
3 of 15
© 0 EX >
a —
-05 | ES d m
Il 5 =
S LE L 1 1 | j 3
=] 0.50 05 1
T T T
— 1 E 7|
g «T 15
© à (T =
ci NA
Il -0.5 I i 5
o " 1 1 1 1 q 3
-| -05. 0.05 1
[ T. T T T
1 E =
à | 15
e ©
ga , NM =
Il -05 d =
© x | B
| | L À
-1-05 0 05 1
Figure 5: Siemens - star at various steps of image sharp-
ness (Az = 15um, on = lgr, SNR= 255). left: test
image, right: histogram of edges, resolving power of opti-
mal edge detector.
The method gives reasonable results: For each test image,
the histogram of edges is a circle with the correct radius
1/0, being the reciprocal width o of the point spread func-
tion used to generate the image.
Test on noisy data. To test the sensitivity of the algo-
rithm with respect to image noise the Siemens star o. =
T
0.6
0.4
0.2
T
dd
02
-04 Fr
-06 |, !
-0.6-04-02 0 0.2 0.4 0.6
83 — 2[gr]
un TI
TE
0.6 - =
0.4
0.2
dd
1
4[gr]
Ne SH Ske Ske SK
=0.2 [=
—0.4
060 iud | =
—0.6-0.4-0.2 0 0.2 0.4 0.6
213
E
rn OT
On
T T T
0.6
04r
02r
—0.2 =
—0.4 FE
“08, Lu l
—0.6-0.4-0.2 0 0.2 0.4 0.6
unu/ 6
= T T T T T
0.6
04 - Shes -
A * ey
02 > p 7
dd
16[gr|
>
03880
$es|s*
2b X d uu
—0.4 tei +
O6 xd Lu 1]
—0.6-0.4-0.2 0 0.2 0.4 0.6
Op =
uu 6
ETT ! T
0.6 -
0.4 - T) z
02r
0
3025 ND
-04 Fr
-0.6
On = 32|gr]
uuy| 6-d*l
[1 31 4-3 55
—0.6-0.4-0.2 0. 0.2 0.4 0.6
Figure 6: Siemens star with o — 2.8 pel at various steps of
image noise (SNR=128, 64, 32, 16, 8).
2.8 pel from fig. 5 was speckled with Gaussian noise, the
noise variance being o2.
The results in fig. 6 show that the method is quite robust
with respect to image noise. Note that the slightly decreas-
ing resolving power of the ideal edge from the first to the
last row is caused by the increasing image noise.
3.2 Results on real data
Real data with artificial blur. The method was also tes-
ted on a real image of the MIT building which was grad-
ually blurred by convolution with Gaussian filters of in-
creasing filter width (cf. fig 7).
We see that the method seems to yield correct results. In
almost each histogram of edges the ellipse containing all
points is elongated, indicating anisotropy of the image sharp-
ness for the given image.
Aerial image with various sharpness. Finally, the me-
thod was applied to digitized versions of an aerial image
(cf. fig. 8, top row) scanned three times with a pixel
size of 7um. Various image sharpness has been realized
physically by imposing layers of transparencies between
the original and the scanner platform, thus exploiting the
limited depth of view of the optical system of the scanner.
A - 209
