oppo- 
circu- 
iter of 
bution 
direc- 
nce, if 
dging 
‘om to 
rmine 
vs, Le. 
linear 
|, then 
> The 
much 
ve no- 
htness 
some 
re for 
round 
angle 
flector 
or alti- 
ze and 
water 
mage. 
use a 
ng the 
psition 
€ usu- 
ff. 
above. 
is not 
emati- 
ımma- 
'ection 
ion of 
he im- 
>, then 
ar. 
odging 
algorithms. The first way is to identify all factors 
involved in causing unbalance and then quantitatively 
model them. One may be able to quantitatively determine 
some factors, but it will be very difficult to determine all 
of them, because many random factors are also involved. 
The first way has its theoretical foundation, but practi- 
cally it is very hard to implement. The other method is to 
establish a generic polynomial model for all factors, ran- 
dom or not. First order polynomial represents a linear 
model while the higher order polynomial describes non- 
linear model. The generic solution method is easy to be 
implemented and has been proven to be effective. 
ImageDodge developed at PSI uses the generic solution 
method and has been successfully used to process hun- 
dreds of orthoimages. 
Because of the complexity of radiometric unbalance, it is 
very hard to find a set of parameters which is good for 
all images. For example, we often see that orthoimages 
for one project have different unbalance patterns, either 
because original aerial photos were taken at different 
times, or because orthoimages were generated out of dif- 
ferent areas of original photos. In this kind of situation, 
images have to be processed on a one by one basis, in- 
stead of being processed in batch mode. Also, to process 
one image, you may have to try several times to get the 
best adjustment for brightness and contrast. 
Two pairs of orthoimages are shown at the end of this 
paper. Figure 1. Shows a pair of images with unbalanced 
brightness and contrast. There is a mismatch between left 
image and right image. The left side of each of the two 
image is lighter than the right side. Figure 2. are the two 
same images after being dodged by ImageDodge. In this 
figure, it is hard to tell where the edges of the two images 
are. 
4. CONCLUSION 
With the increase production and use of orthoimages, 
people now are paying more attention than before to im- 
age radiometric quality. Because some problems which 
are not that noticeable in individual orthoimages are very 
apparent when orthoimages are mosaiced together. Ra- 
diometric unbalance in brightness and contrast within 
individual images is the biggest problem for image qual- 
ity. Existing image processing software can reduce radio- 
metric differences among images very effectively if the 
images themselves are internally balanced. But they can 
do little for internally unbalanced images. ImageDodge 
developed at PSI has been successfully used to process 
hundreds of orthoimages. But because the complexity of 
radiometric unbalance, it now runs only under a manual 
mode. The future research and development at PSI will 
be focusing on investigating whether and how the dodg- 
ing process can be automated, especially on finding a 
205 
way that can automatically and quantitatively determine 
the amount of distortion. 
REFERENCE 
Hussain, Z., 1991. Digital Image Processing. Ellis 
Horwood Limited, England, pp. 213-218. 
Kiefer, L., 1994. Remote Sensing And Image Interpreta- 
tion. Third Edtion. John Wiley & Sons, Inc., USA, pp. 
13-20. 
Knabenschuh, M., 1995. Radiometric correction of 
orthophoto data. International Journal for Geomatics, 
Vol.9, 10, pp.72-74. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996