> Recently, I started building mesh2s following Earth's curvature from
> SRTM Earth elevation data (3 arcseconds per data point)... when
> rendering those mesh2s, I found some strange artifacts (gaps between
> triangles, see image on p.b.i.) which probably resulted from floating
> point inaccuracy (the scale I used is 1 POV unit = 1 kilometre).

Are your adjacent triangles referencing the same vertex, or are you 
defining the coordinates repeatedly for each triangle?

> As Christoph Hormann, as he writes on his website, apparently uses a
> kind of (self-written?) conversion tool, I thought that converting my
> mesh2s to isosurfaces possible could solve the gaps problem...

It seems unlikely the isosurface rendering code would provide better 
accuracy than the triangle rendering code, if indeed you are hitting 
some limitation with POV.  IIRC the triangle intersection code is only a 
few calculations, whereas the isosurface intersection code is much more 
complex (which means more calculations and more possibility of floating 
point errors building up).

Post a reply to this message