F X)-Épo/X)V, TX -u) (14) 
Th JX) in Eq.(14) can be calculated by Eq.(9). And 
Jes, 
$ mw 
unknown parameters V j and I’ ; can be calculated through 
least square method. It needs to meet Eq.(15) 
(15) 
  
:-Yr-F a 
Kain et al. (1998) made some improvements by putting 
X, and Y, together as a Joint vector z AK ‚The 
optimal transformation function is 
= y AR x 16 
FAX Zah, Er pt a} 49 
where u ; ur "= the mean vector of the source texture 
J J 
and destination texture in class j 
ST = the covariance of source texture in class j 
j 
S — the cross covariance of the source texture and 
j 
destination texture in class j. 
The model parameters of Z, are 
lu (17) 
H, 
y-2 E am 
SN 
The method proposed by Kain can make the GMM more 
reasonable but will increase the amount of calculation. 
After transforming, we use Eq.(19) to evaluate the accuracy 
p FO -Y| (19) 
  
4. EXPERIMENTAL RESULTS AND CONCLUSIONS 
This paper uses QuickBird imagery and SAR imagery in the 
same area with spatial resolution of 1m. figure4 and figure6 
show the Gabor filtered outputs of QuickBird and SAR imagery. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
    
   
Xtreme qur 
  
     
© ee 
Figure 4. Textures of QuickBird imagery with different 
orientations(a to f stand the texture images of g — 7/ to 0=x 
ith Tc 
with every rA )