2 
3 
| 
Misclassified pixels 
  
-2000 — 
Class 5 
Class 6 
Class 7 
Class 9 
  
  
Compression ratio 
| ' | 
Figure 4. Misclassified Pixels for Some 
Selective Classes of E. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Class | Ex.GIS | Ex9% Ex11% | Ex12% | Ex13% | Ex14% 
1 4875 5487 5574 5618 5655 5629 
2 7322 6815 7006 6821 6741 6834 
3 4080 4705 4601 4763 5050 4829 
4 5988 5164 5285 5161 5074 5073 
5 4236 5014 4836 5159 5009 5158 
6 5442 5434 5373 5790 5472 5348 
% 6636 6519 516119 4838 6230 5834 
8 9666 9466 9643 9759 9102 9732 
9 13689 13367 13558 14082 13615 13535 
10 3342 3563 3536 3543 3586 3564 
  
  
  
  
Table 2. Number of Pixels of Original 
Ex.GISmap and Five Compressed GISmaps. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
3 
| 
lend 
  
  
0 
o 
‚X -— 
a 
© 
o 
= 
a d 7 Compression Ratio 
© 
o : | | : | : I : 1 
2 D 10 11 
2 12 13 14 
-1000 
Figure 5. Misclassified Pixels for Some 
Selective Classes of Ex. 
It can be concluded that image compression 
using JPEG is a scene dependent process, and 
the result of one scene can not be generalized 
for all scenes. Another remark is that features 
which are very distinguished in nature and have 
sharp edges with their surrounding areas may 
survive high rate of compression. Such a 
characteristic may have valuable applications in 
digital mapping. 
Table 3 and Table 4 summarize the amount of 
data reduction at different rates of compression 
as exercised in experiements E and Ex. It can be 
pointed out that more than 700,000 bytes can be 
reduced for the three band TM image with a 
maximum of 4.5% misclassified pixels ( or 
loss of information) out of 65,536 pixels. 
  
  
  
  
  
Image Size Reduction 
E.lan 786944 0 
E8% 100033 686911 
E10% 78477 708467 
E12% 66284 720660 
  
  
  
  
  
Table 3. Amount of Reduction in Storage 
for Experiment E (in bytes) 
27