257 
In: Wagner W., Szekely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
tuation: We 
-an be from 
olution, or a 
by a multi- 
iput images, 
hich can be 
: at different 
captured at 
. Our aim is 
a dynamic 
ed in Figure 
nes that we 
iages and a 
ge relates to 
sess several 
ngths while 
•all image is 
generalized 
^constructed 
rmbining all 
nultiplexing 
wavelengths 
taken from 
ation of the 
te LR image 
The Fourier 
new image. 
1 image, we 
the Fourier 
erse Fourier 
the various 
l images by 
; rest of the 
impose the 
ic algorithm 
obtained in 
threshold. 
'rom regions 
e strength of 
directly on 
:s mainly on 
the a priori 
some of the 
constraints 
results from 
I | Actions performed from first cycle and on. 
Figure 1. Flow chart of the proposed generalized super 
resolving approach. 
EXPERIMENTAL VERIFICATION 
In this section the authors present two types of experimental 
results, both coming from airborne camera capturing urban 
area. In Figures 2(a) and 2(b) we present two experimental 
results. In both, the upper row contains from left to right: a high 
resolution image, the same image taken at a quarter of the 
resolution and the obtained reconstruction. The cross section is 
seen in the lower row. Red corresponds to the reconstruction 
image, green is the quarter resolution image, and blue is the half 
resolution image used for comparison. The reconstructed image 
obtained from the multiplexing of the high resolution image and 
the quarter resolution image follows the cross section of the half 
resolution image very well while having an improved contrast 
by 16.7% in Figure 2(a) and by 12.3% in Figure 2(b). The 
correlation between the reconstructed image and the HR image 
is 98.2% and the mean square error is less than 0.8%, for the 
later test case. 
In the second experiment presented in Figure 3, we obtain the 
photograph of the same scenery using two different 
wavelengths. We assume that in the shorter wavelength image 
we succeeded to obtain only the lower right quarter of the 
image. By the same iterative procedure (but using different 
parameters) we generate a new image (result) containing high 
resolution data even outside the imposed data region. In the 
lower row of Figure 3 one may see the reconstructed result 
(left) and its cross section (right). An improvement of more than 
10% in the contrast is obtained when comparing the 
reconstruction results (red) and the original image at low 
resolution (longer wavelength). 
High resolution 
Quarter 
resolution 
reconstructed 
16.7% improvement in contrast Zoom of cross 
Note that the 
reconstruction follows 
the half-resolution 
image quite well (with 
improved contrast) 
especially when 
compared to quarter 
resolution image. 
High resolution 
Quarter 
resolution 
reconstructed 
12.3% improvement in contrast Zoom of cross section 
Again the reconstruction 
follows the half-resolution 
image quite well (with 
improved contrast) 
especially when 
compared to quarter 
resolution image. 
Figure 2. Experimental results using airborne camera images, 
(a). Captured image #1. (b). Captured image #2. In both cases, 
the upper row contains from left to right: a high resolution 
image, same image at quarter of resolution, and the 
reconstruction results. The cross section is seen in the lower 
row. Red is the reconstruction, green is the quarter resolution 
image, and blue is the half resolution image used for 
comparison. The reconstructed image obtained from 
multiplexing of the high resolution image and quarter resolution 
image follows the cross section of the half resolution image 
very well while having an improved contrast (improvement by 
16.7% in (a), and 12.3% in (b)).