Christoph Hormann <chr### [at] gmxde> wrote:
> Russell Towle wrote:
> >
> > DEM data is better handled by POV's "height field" function, which creates a
> > mesh of triangles on the fly.
>
> In this general form this is certainly wrong.  Using an isosurface is
> more memory efficient and you can get much higher quality results (no
> mesh artefacts).  In most cases an isosurface will render slower than a
> heightfield but often this is not the most important factor.

I have no experience with the isosurface method.

It is true that the mesh of triangles created in the POV height field
becomes visible when one zooms in close enough. This is rarely a problem
for me; I use the highest-resolution DEM data commonly available in the
U.S., the 30-meter DEMs, which span 7.5 minutes of latitude by 7.5 minutes
of longitude. I will often merge larger numbers of these DEM quadrangles
before rendering in POV. For instance, recently I've been using a merged
set of thirty DEMs. This gives me a patch of terrain of something like
twenty or thirty miles square (though it is not really "square").

Typically, I never see any sign of the mesh itself. I am interested in
landscape-scale geomorphology and geology: the incision of canyons, here in
California's Sierra Nevada, the difference between glacially-eroded and
stream-eroded canyons, the difference between canyons incised into granite
versus canyons incised into metavolcanic rock versus canyons incised into
metasedimentary rock.

These differences are often astoundingly visible using the merged 30-meter
DEMs.

So, for my own purposes, the height field works quite well.

Russell Towle

Post a reply to this message