Lerma, Jose Luis 
  
The texture filtering generated the texture images to be used in conjunction of the four spectral bands. In fact, only the 
most representative parameters (for these monument) were selected for the red spectral band applying the GLCM, the 
mean and the homogeneity. In Fig. 2a appears the homogeneity image for a window of 3x3, whereas in Fig. 2b and 2c 
appear the mean and homogeneity images for a window of 50x50. Finally, the whole set of seven images was extracted 
in order to have the same physical size (resolution). 
    
(b) 
Figure 2. Textural images: (a) homogeneity (kernel 3x3), (b) mean (kernel 50x50), (c) homogeneity (kernel 50x50) 
4 SPECTRAL AND TEXTURAL CLASSIFICATIONS. RESULTS 
The informational classes used for the classification were defined in the selection process and proved to be 
distinguishable. Three sets of extracted images were classified by means of the standard supervised rule of maximum 
likelihood. All the classified images were evaluated by means of visual inspection in the field, and qualitatively 
analyzed. The first set was only integrated by the visible bands; the result is shown in Fig. 3a. The second set was 
integrated by the visible bands and the near-infrared band; the result appears in Fig. 3b. The last set used the whole set, 
the three visible bands, the near-infrared band and the three textural bands; the result appears in Fig. 3c. 
   
@ TT Ee E t i NC: 
Figure 3. Classified images: (a) spectrally (G,B,R bands), (b) spectrally (G,B,R,IR bands), 
(c) spectrally (G,B,R,IR bands) and texturally (3 bands, Fig. 2) 
The visible spectrally classified image, Fig. 3a, shows only one spectral class properly recognized, clean cement mortar. 
Most of the other classes have too many defect or excess errors. For example, flaking is neither well placed nor well 
distributed; polished limestone is rather confused, mainly over the other kind of limestones, clean and dirty; dark 
cement mortar offers flooding over washed limestone. On the other hand, washed limestone only fits the area and 
follows the traces; and finally, dark limestone seems well-classified although it has some expansion over polished 
limestone. 
Regarding the second classified image, Fig. 3b, there is a better discrimination between cement mortar-limestone, as 
well as among flaking—washed & washed limestones. Unfortunately, this classification still has got a general confusion 
among polished limestone and washed & washed limestones. 
  
482 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 
Finally 
The co 
there 1 
Partict 
8 CC 
The re 
the ch 
recogr 
The u! 
identi 
discrit 
(much 
proces 
REFI 
JM. ( 
Intern 
R. M. 
J. He 
identi 
Photo 
Koda 
JL. 
XI C 
J. L. 
Reco 
Patte 
J. 1. 
Text 
http: 
S.R 
Phot 
work