Full text: Technical Commission III (B3)

    
   
  
   
   
   
  
   
   
      
    
    
     
    
    
  
    
    
   
    
    
    
    
    
  
   
   
  
   
       
        
            
  
t, the brightness, 
and non-shadow 
nal image band. 
ling indices are 
le 1 is the result. 
e changed, when 
oval processing. 
id charts of the 
Table.1 and the 
the brightness 
ecreases sharply 
e shadow region 
more short-wave 
gradients in the 
f decrease in the 
ased on Color 
shadow regions, 
yal with p=2, the 
improved mostly 
he value in non- 
ed too, but not 
show that the 
of three bands 
is a value from 
is 2. And when 
verage gradients 
region. 
region color is 
olor error exists 
oG algorithm. It 
1e best result of 
nent. 
removal 
88 
  
47 
  
p 
D 46». 00! C2 
eio cio: do 
  
0 5 10 15 
p 
  
  
brightness 
c 
contrast 
  
  
  
  
Figure 3. Change trends of brightness, contrast and average gradients when p value from 1-10 in true color aerial image shadow 
removal 
5.2 Shadow removal experiment with color infrared aerial 
images 
The test image is a color infrared aerial image of Shanghai from 
1998; the resolution is 20cm/pixel. In the experiment, the 
original image was reduced to 704x734pixel; the resolution is 
0.4m/pixel. The Figure 4(a) is the original image. The Figure 
4(b) is the shadow detection result; the black area is shadow 
region. Figure 4(c) is the shadow removal image when p is 2. 
     
  
(a) (b) Cute) pen. 
Figure 4. Color infrared aerial image (a) the shadow detection 
result (b) and the results of shadow removal (c) 
Using the same method as 5.1, the brightness, contrast and 
average gradients are calculated separately in the shadow and 
non-shadow regions. Table 2 is the result. Figure 5 includes 
three statistical trend charts of the shadow region, when p is 
from | to 10 in shadow removal processing. 
From Table 2 and the Figure 5, we draw the following 
conclusions: 
(1) Compared to true color aerial images, the brightness 
decreases more sharply in shadow region, especially in red and 
infrared band. The contrast and average gradients decreases 
similarly; the degree of decrease is the same in all three bands, 
and the result is the average gradients decreases much more 
than in the true color image experiment. 
(2)When p is 1,or 2, the SoG shadow removal algorithm can 
recover the brightness, contrast and average gradients of 
shadow region in color infrared aerial images, especially for the 
brightness in red and infrared band. The values are close to non- 
shadow region after shadow removal. 
(3)From Figure 5, the change trend lines show that the 
brightness, contrast and average gradients in red and green 
bands achieve maximum value when shadows are removed 
using p=2. In the infrared band the three statistical 
characteristics achieve maximum when p-1; the three values 
decrease a little when p-2. Infrared band measured values 
sharply drop when p is from 2 to 10. When p is 2, the shadow 
contrast and average gradients have a good agreement with non- 
shadow. Considering the brightness, contrast, and average 
gradients improvements in the shadow region across red, green, 
and infrared bands, the color infrared aerial image achieves the 
best overall result of shadow removal when p values 2 in this 
experiment. 
(4)The visual effect shows that when p is 1, 2, or 3, the shadow 
region color is closer to non-shadow region after the shadow 
removal; the effect is better than the resulting image when p is a 
large value. There is no visible color error after the shadow 
removal processing with the SoG algorithm. It follows from this 
experiment that for color infrared aerial images, best results for 
shadow removal are obtained when p equals 2. 
Table 2. The brightness, contrast, and average gradients of the color infrared aerial image before and after shadow removal 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
brightness contrast average gradients 
NIR R G NIR R G NIR R G 
Shadow region | 12.0784 | 13.6279 | 34.3757 | 6.7721 | 7.3936 |11.2968| 0.485 | 0.5063 | 0.8062 
1 76.8154 | 80.51 |100.6389| 43.0678 | 43.6788 | 33.0724 | 3.0843 | 2.9913 | 2.3602 
2 75.9624 | 82.9082 |102.5798| 42.5895 | 44.9799 | 33.7102 | 3.05 3.0804 | 2.4058 
3 | 64.8578] 73.6396 [101.9624] 36.3635 | 39.9515 | 33.5074 | 2.6041 | 2.736 | 2.3913 
4 50.184 | 58.018 | 98.5613 | 28.1365 | 31.4763 | 32.3897 | 2.015 | 2.1556 | 2.3115 
Pg 5. |39.5395 | 45.8608 | 92.5138 | 22.1684 | 24.8808 | 30.4023 | 1.5876 | 1.7039 | 2.1697 
6 32.1539 | 38.1539 | 84.9458 | 18.3934 | 20.6995 | 27.9153 | 1.3172 | 1.4176 | 1.9922 
7 28.3917 | 33.1515 | 77.4869 | 15.9183 | 17.9856 | 25.4641 1.14 1.2317 |]. 1.8173 
8 25.3295 | 29.7025 | 71.0948 | 14.2014 1 16.1144 | 23.3635 | 1.017 | 1.1036 | 1.6674 
9 23.105 | 27.1962 | 65.9188 | 12.9542 1 14.7547 | 21.6626 | 0.9277 | 1.0104 | 1.546 
10 |21.4306 | 25.2988 | 61.7694 | 12.0154 | 13.7253 | 20.299 | 0.8605 0.94 1.4486 
unshadow region | 76.8126 | 80.5071 |100.6347| 41.0305 | 49.1429 | 39.1242 | 3.358 | 3.4974 | 3.1397 
  
  
  
  
  
  
  
     
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.