anbul 2004 
Mean 
ctrum 
r values of 
nd moment 
ment. 
NR of each 
zl H3 
v-- 121 
  
1t10 
  
  
16 
n ratio 
  
  
1 ratio 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
component of RGB color space was similar to that on PSNR of 
Value of HSV color space, we do not show graphs of PSNRs of 
RGB components. 
Furthermore, graphs of the effect of changing compression ratio 
on image quality measures are illustrated in the compression 
ratio range from 1 to 16 in our judgment that highly compressed 
images, that is, low quality images are not suitable for most of 
GIS application. 
(A) Figures 5 and 6 show the effect of changing the IJG quality 
setting on the compression ratio. The compression ratio is 
defined as the ratio of the number of bytes of the source 
image before compression to the number of bytes of the 
compressed image data. 
The compression ratio of the compressed image data of 
each color component (Red, Green and Blue) in Type RGB 
is shown in Figure 5. Figure 6 shows the compression 
ratios of Blocks C2 and H3 as well. 
(B) Figures 7, 8 and 9 show the effect of changing 
compression ratio on PSNRs of Hue, Saturation and Value 
of HSV color space respectively. 
(C) There were no significant differences among the effects of 
changing compression ratio on the texture measures of Red, 
Green and Blue of RGB color space. We provide graphs 
of texture measures of Green against the compression ratio 
in Figures 10 to 13. 
Each texture measure is illustrated in the ratio of the value 
of the reconstructed image to that of the source image. A 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
       
  
  
  
1.004 du ei 
=} 3 
S 1.000|. A RO AN 
3 | 
> 
5 A 
= 0.996! 5411 
= te PS Thine 
^ € - 53434 
3 AL CUN OK 
20.992 oe ene 2] H3 
us RGB 
cz 
0.988 
0 4 8 12 16 
Compression ratio 
Figure 10. Standard deviation vs. compression ratio 
1.04 E e I 
1.00 ee RGB 
S 
2 
© y 
e 0.96 
= 
= 
© 
ey 13 
.C 
0 4 8 12 16 
Compression ratio 
Figure 11. GLDV contrast vs. compression ratio 
ratio of a texture measure closer to 1.0 indicates that the 
reconstructed image is closer to the source image in texture 
feature. 
Figure 10 shows the effect of changing compression ratio 
on standard deviation. Figures 11 and 12 show the effect 
of changing compression ratio on GLDV contrast and 
GLDV angular second moment respectively. Figure 13 
shows the effect of changing compression ratio on mean 
spatial frequency of Fourier power spectrum. 
3.5 Discussion 
There were significant differences in the compression ratio 
against the IJG quality setting among three Types S444, S411 
and RGB as shown in Figures 5 and 6. While Type S411 
provided the most compressed image, Type RGB with no color 
space conversion had the least compression efficiency. 
Differences in the compression ratio among each color 
component (Red, Green and Blue) in Type RGB were rather 
small. 
As the IJG quality setting decreases, differences in the 
compression ratio among Types S444, S411 and RGB increases. 
Furthermore, the compression ratios in all types increase rapidly 
with decreasing the IJG quality setting range above 20 as shown 
in Figure 5. 
There are differences in the compression ratio among Type 
  
  
   
   
  
  
  
  
e 
—— 
5 1.5 7 RGB 
c 442 f 
pa ene Bui, y 
o "m 
E —— RGB zt 
= 13 ee di 
S 2 
Q P 
BA Hs 2 
S 
= 
= 
< 
> 
Cy 
kl 
© 
  
  
  
  
  
  
  
  
  
  
  
  
1.0 
0:9 
0 4 8 12 16 
Compression ratio 
Figure 12. GLDV angular second moment 
vs. compression ratio 
1.04 eS 444) 
S41] 
23 
9 
= 1.00 
= 
= 
E s 
= 0.96 
= 
2 : 
S Dm rl 2K 
© > “gr wid 
© 0.92 ho C" H3 
RGE = 
0.88 
0 4 8 12 16 
Compression ratio 
Figure 13. Mean spatial frequency vs. compression ratio