Wu,

Nice work, both on cline and superell!  Those are some impressive
pictures of the superell raytraces.  It's probably worthwhile to take
a couple minutes to upload a few pictures of the cline as well - you
can show the implicit and the brep raytraces from a couple different
views, and make a note that the differences are expected and due to
the view-dependent nature of the cline primitive.

I've been thinking a little about the dsp - I suspect the main reason
it is so slow is the (very) large number of control points that result
from inserting all of the grid points into the bezier surface.  There
may be a way to get a reasonable fit without such a large control
point count:

1. the first thing to try is to identify all local maximum and minimum
points in the dsp grid, and insert just those into the bezier surface.
 I.e. for each grid point, check that all of its surrounding points
are either all greater or all less than that point.  That
categorization can be made in a single pass over the grid - the
comparison of the two heights should be computationally cheap, so
repeating checks (i.e. former neighbor point is checked again against
the old center point not that the neighbor is the center) as you
raster your way across the grid shouldn't be a big deal.

2. if that doesn't produce an accurate enough surface, the next
refinement is to make another pass and insert all points which have a
"large" delta height with one or more of their neighboring points.
The heuristic here would be to find the min and max heights in the dsp
grid, and have some "threshold" value for the deltas based on the
relative size of the current max-neighbor delta and the overall max
delta.

With any luck, some form of those two selection methods can produce an
accurate enough smooth surface without the huge control point count.

Cheers,
CY

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
BRL-CAD Developer mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/brlcad-devel

Reply via email to