CGAL users discussion list

[Qianqian Fang <>]

The attached zip file contains a 3D binary image stored in INR format.
I found that CGAL surface mesher produced strange results with this 
dataset.

By using a threadhold value=0.5, bounding sphere center=(122.0, 82.0, 
93.0),
radius^2=32258, and CGAL::Surface_mesh_default_criteria_3<Tr> 
criteria(30,4,4),
I run the mesh_a_3d_gray_image example with this data, and unfortunately,
the program had ran forever and exhausted the system memory.

Then I modified include/CGAL/Surface_mesher/Surface_mesher.h
void refine_mesh(Visitor visitor) function and set a termination condition
when tr.number_of_vertices() reaches certain number. If I set this
number be 40000, I got some reasonable surface mesh output.
However, if I make the limit bigger, say 80000, the program will return
a mesh with similar mesh density. Plotting the nodes, I found that there
are huge number of them aggregate around coordinates (39   93   86);
a plot for all the nodes between two narrow planes x=38.8 and x=39.6
is shown in the attached figure (note the x-axis span is indeed very small).

From the plot, it looks like the mesh refinement was trapped near this 
mentioned
position and new points were added in a straight line fashion.

I am curious if this is an expected behavior. I know setting the angular 
bound
may cause some troubles for certain structures, but I don't know if this
straight-line trapping is one of those.

your comments are appreciated.

Qianqian

Attachment: pre_extract.zip
Description: Binary data