569 
RESEARCH ON THE BUILDING 
SHADOW EXTRACTION AND ELIMINATION METHOD 
GUO Hai-tao*, Zhang Yan, Lu Jun, Jin Guo-wang 
Zhengzhou Institute of Surveying and Mapping ,66 Mid-Longhai Road, Zhengzhou, China, 450052 
-ht7776@163.com 
Commission III, III/5 
KEY WORDS: Building, Shadow Extraction, Shadow Removal, Multiple Constraints, Edge Detection 
ABSTRACT: 
This paper first generalizes the significance and present research situation of the building shadow extraction and elimination method. 
Then, an approach of automatic extracting shadows of building from remote sensing image is proposed, which based on the method 
that grey level statistics is done in one side of the structure lines. The approach, which combines constraints of grey level, geometry 
and context, and others assistant information, realizes the shadow extraction of buildings. Finally the IKONOS images are 
experimented to extract shadows of building. Results of the experiment show that this approach is effective and robust. 
1. INTRODUCTION 
Shadow in remote sensing image is conflict information in 
computer image processing. On the one hand, it can reduce the 
successful rate of edge extraction, object recognition, image 
matching, change detection and other processing for the 
corresponding ground objects in the shadow (Wang et al, 2002; 
Yan 1 i et al, 2004). On the other hand, it can produce a great 
deal of useful information about shape, relative position, 
surface character and other characters of the object generating 
shadow (Barnes et al, 1999; Ortega et al, 1998; Yang et 
al,2002). It is a necessary step to eliminate shadow and restore 
the scenes in the shadow area before performing object 
recognition and image matching tasks for the shadow area. 
Therefore, it is of vital importance and keen interest to discuss 
and investigate the shadow extraction and elimination 
technique. 
Research on the shadow extraction is few in remote sensing 
field. Only the research on the tall object recognition (Lin et 
al, 1998) shadow elimination and shadow compensation (Wang 
et al,2002; Yan li et al, 2004; Amani et al,2004) are carried out. 
One kind of shadow extraction methods is carried through 
detection of histogram peak value at present because the grey 
value of shadow is very low in all visible wave bands (Nagao 
et al, 1979). But this method can not deal with some complex 
situations. Statistics of pixel grey values along the structure 
line is the second kind of shadow extraction methods. But the 
extraction result of this method depends on the credibility of 
shadow edge extraction result. When the credibility of shadow 
edge extraction result is low, shadow extraction will not 
achieve an accurate result. The building shadow extraction 
under the multi-constraints is another class of method (Amani 
et al,2004; Rau et al, 2000; Babu et al, 2004 ). These 
constraints consist of the grey condition, the geometrical 
condition, context condition and other auxiliary information 
condition. But this method is difficult to carry out and its 
algorithm is very complex. 
Shadow elimination or shadow compensation is often carried 
out after shadow extraction. The methods of shadow 
elimination or shadow compensation mainly conclude 
adjusting light and contrast (Rau et al.2000), using illuminating 
model and contextual texture consistency analysis (Amani et 
al,2004). Simply adjust the brightness and contrast of the 
image can approve the image quality, but shadow elimination 
effect usually is not very good. There are methods restoring the 
shadows in satellite images using illuminating model. However 
it can not be copied to aerial images, because camera height for 
aerial imagery is much lower than satellite imagery, the 
material feature of the ground object affects to its intensity in 
the image. When the texture information of shadow area is not 
consistent with that of its neighborhood, the method contextual 
texture consistency analysis can not deal with. 
In the paper, a comprehensive analysis and discussion is 
carried out on the present commonly used building shadow 
extraction methods. Then based on the statistics shadow 
detection method that detects the shadow area through the 
statistical analysis and calculation of pixel grey values along 
the structure line, a new building shadow extraction method 
under the multilateral constraints is put forward. The new 
method realizes the shadow extraction by combining the grey 
constraint, the geometrical constraint, the context constraint 
and other auxiliary information. 
Shadow elimination method is also studied in the paper. First, 
the pixel grey value of the interior area of the shadow and that 
of the surrounding region along the shadow edge are 
statistically calculated and analyzed (Yan li et al, 2004). And 
then the transform parameter between the interior region and 
the surrounding area are reckoned through gamma transform. 
Finally the shadow is eliminated through gamma transform 
method and the scenes in the shadow are restored. 
* This paper is financed by the item of Institute of Surveying and Mapping Zhengzhou.