Full text: Proceedings, XXth congress (Part 5)

  
  
  
  
   
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
2.5 Texture synthesis 
As already pointed out, many elements are damaged in the 
sense that parts have broken off. But the surfaces are also 
eroded. As a result, the original textures and colours have 
disappeared. Even if 3D shapes arc retrieved and restored, the 
texture to cover them with in order to restore the full, original 
appearance cannot be obtained from their available imagery. We 
have in this case wetted the least eroded parts for some 
materials to mimic the effect of polishing, and have taken 
images of undamaged surfaces of the same materials but not 
found at the site for the remainder. Based on such images, 
texture models were generated and used to produce arbitrary 
amounts of similarly looking texture (Zalesny, 2001). A total of 
17 different building materials were used in this building. Fig. 7 
shows example images for some materials, as well as synthetic 
texture patches based on these examples. The necessary 
amounts of synthetic texture were then generated for each type, 
and with the necessary patch shapes to cover the elements to 
which they apply. 
Rather than describing these 3D modeling and texture synthesis 
techniques any further, we here describe the additional steps 
needed to put the components together into the final model. 
2.6 Assembling the components 
With blueprints (figure 8, top and middle left) as guidance, we 
put all components together within Maya. 
Of course, some elements couldn't be scanned, because they are 
still missing or have been destroyed. Moreover, even if the 3D 
capture technology is easy to use. scanning every block from all 
sides is still too expensive (pieces would have had to be moved 
with a crane). Therefore, the lacking components have to be 
modeled by hand using common modeling methods like: curve 
drawing and extrusion (translational sweep), solid modeling and 
booleans, cutting tools, bevel tools, deformation etc. Further 
reading on these methods can be found in numerous books (e.g. 
Mortenson, 1985), software manuals (e.g. Alias, 2004) and 
papers (e.g. Sederberg, 1986). Fig. 8 shows in the middle on the 
right the hand-modeled back wall and the basin of the 
Nymphaeum. The image on the bottom shows the fully 
assembled monument. 
To simplify the manual modeling process, 3D scans can also be 
used as a shape outline one can constantly refer to. The efficient 
and cost-effective nature of the ‘shape-from-stills’ pipeline 
allows for this luxury. In addition, such coarse scans are much 
easier to make than technical drawings for every construction 
layer. Fig. 9 shows at the top the aligned scans and in the 
middle the reconstructed model. Having both models, difference 
functions can be calculated automatically and visualised on the 
models, for example fig. 9 on the bottom: the grey areas 
highlight the parts that had to be extrapolated from other 
archaeological data, mainly from other pieces of the 
nymphaeum. Such representation gives an idea of local 
uncertainty about the model. 
  
travertine limestone with limestone limestone marble from “white” limestone 
deposit chert nodules breccia 1 block with Dokimeion crystallized breccia 2 
pink styloliths limestone 
(type 10) (type 3) (type 8) (type 6) (type 8) 
(type 1c) (type 5b) 
Figure 7: Example images of building materials (top) and synthetic textures (bottom) based on 
models extracted from these examples. 
   
  
  
   
   
   
  
   
  
   
   
  
   
   
   
  
  
   
   
   
   
  
   
  
   
  
  
   
  
   
   
  
    
  
   
  
   
   
  
  
  
 
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.