house 
Ur 
a.gov 
html 
CS, 
<m net 
ject. 
Vol. 
vide 
‚stem 
A.html 
ISBN: 
DIGITAL IMAGE DODGING IN DIGITAL ORTHOIMAGE PRODUCTION 
Zheng Wang 
Photogrammetry/Image Processing Specialist 
Photo Science, Inc. 
Gaithersburg, MD 20878 
U.S.A. 
Commission I, Working Group 1 
KEY WORDS: Digital Orthoimage, Image Dodging, Digital Image Processing, Image Radiometric Quality, Image 
Mosaic. 
ABSTRACT 
With the increase production and use of digital orthoimages, the radiometric quality of orthoimages has caught more 
attention then before. Digital orthoimages are often found to be distorted or unbalanced in brightness and/or contrast 
within individual images and/or among images. This is true especially for color orthoimages. We often see digital 
orthoimages with unbalance problems, such as fall-off of brightness in image corners, or one side of an image is lighter 
or darker than the opposite side of the image, or the contrast at the top of an image is different from the contrast at the 
bottom of the image. When such kinds of orthoimages are put together or mosaiced together, the radiometric 
mismatches along image edges will show up. The differences in brightness and/or contrast among neighboring 
orthoimages often make one part of an object in one orthoimage look totally different from the rest part of the object in 
a neighboring orthoimage, which in normal case should look exactly same. The unbalance problems make orthoimages 
not only look bad but also increase the difficulty for image interpretation and feature classification. Existing commercial 
image processing software can do little for the problems, because they always process all image pixels equally without 
any consideration about possible unbalance within the image. A software called ImageDodge was developed at Photo 
Science, Inc. (PSI) and has been effectively used to correct or dodge all the digital orthoimages produced since which 
had unbalanced brightness and/or contrast. This paper discusses the unbalance problems for digital orthoimages, the 
distribution patterns of unbalance and their mathematical modeling. The paper also introduces the algorithm used in 
ImageDodge and shows some orthoimages before and after being dodged. 
1. INTRODUCTION mosaiced image, brightness and/or contrast can change 
abruptly across seamlines and sometimes can make entire 
With the increase production and use of digital mosaiced image look a striped image with light stripes 
orthoimages, the radiometric quality of orthoimages has and dark stripes. The differences in brightness and/or 
caught more attention then before [Knabenschuh, 1995]. contrast among neighboring orthoimages often makes 
The reason is that in many cases, produced orthoimages one part of an object in one orthoimage look totally dif- 
are found to be distorted or unbalanced in brightness ferent from the rest part of the object in a neighboring 
and/or contrast within images and/or among images. This orthoimage, which in normal case should look exactly 
is true especially for color orthoimages. Here an unbal- same. All those problems will make orthoimages not 
anced image is defined as any image which has an un- only look bad but also increase the difficulty for image 
even or non-consistent brightness and/or contrast within interpretation and feature classification. 
the image itself. Most digital orthoimages are generated 
out of images scanned from aerial photography. We of- Unbalanced orthoimages have to be radiometrically cor- 
ten see aerial photography with radiometric problems, rected or dodged. But existing commercial software are 
such as fall-off of brightness in image corners, or one almost helpless for resolving the problems we talk here. 
side of an image is lighter or darker than the opposite The reason is very clear: all of those commercial soft- 
side of the image (see Fig. 1), or the contrast at the top of ware assume that each image being processed is 
an image is different from the contrast at the bottom of radiometrically balanced, i.e. the brightness and contrast 
the image. The problems mentioned above Any within the image are consistent. When they change the 
orthoimage generated out of those kinds of images will brightness or contrast of an image, they change every 
have the same radiometric problems as the original pho- pixel of the image equally without any discrimination. If 
tography have. When such kind of orthoimages are put the original image was unbalanced, then the processed 
together or mosaiced together, the radiometric image will still be unbalanced. Commercial image pro- 
mismatches along image edges will show up. Within the cessing software works well to reduce radiometric differ- 
203 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996