International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
frontal face images then the system can easily match it to the 
database image. 
e 
  
Figure 3: Example of the images from ORL face database 
Tables 1 to 3 shows sample of the results obtained. If the face 
orientation of each sample image does not differ to much form its 
frontal face images then the system can easily match it to the 
database image. However if the sample images shows a high 
variance in term of light illumination and the size of the face itself, 
the accuracy of the system reduces. 
Table 4 shows the overall result for all ten images for level 2 
wavelet decomposition. The system accuracy which is calculated 
is 86.25%. It can be seen that some tests showed 10094 accuracy, 
but image 8 has showed low accuracy due to the factor mention 
previously. Figures 4. 5 and 6 show the performance accuracy of 
the proposed system using different wavelet multiresolution and 
decompositions, also the accuracy and the efficient of the 
proposed method using a different number of orientations. 
Table 1 
1 
2 
3 
4 
5 
6 
7 
8 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Table 2 
el 
22 
e2 3 
24 
2:8 
2 6 
27 
28 
S Acc 
Table 3 
es 
Image 4 
I e4 
4 
e4 
e4 
e4 
e4 Yes 
e4 Yes 
A 62.5% 
Table 4 
Images System Accuracy (%) for 
level 2 Wavelet 
Decomposition 
Imagel 75 
Image2 100 
Image3 100 
Image4 62.5 
Image5 100 
Image 6 87.5 
Image 7 100 
Image 8 37.5 
Image 9 100 
Image 10 100 
Total 86.25 
Accuracy 
  
  
  
  
Accuracy of system according to level of Wavelet 
100 
90 
80 
70 
60 
50 
40 
30 
20 
10 
Accuracy of matching 
Figure 4: Accuracy based on Levels of Wavelet 
Transform 
71.25 
1 2 3 4 
Wavelet Decomposition level