Re: clipping and viewport mapping problem

This is
not enough however: you have ambiguity when an edge crosses the center

Actually, the problem is more philosophical. The problem is in the
definition of our shapes. We have used to define our geometry by using
polygons which share edges. The geometric description _is_ unambiguous.
But when we associate materials, each polygon might contain a different
texture. Additionally, we assume that each polygon owns its edges (and
vertices) and thus that its edges has the same material. When two or
more polygons do this to the same edge in parallel, we have an
ambiguity.

Now, using the normal rule, just what Dave suggested, this ambiguity is
solved in run-time. However, because the solving is based on screen
space orientation of the polygon, the solution is orientation dependent!
Here's a test to see this: render an image, make a 180degree roll with
the camera, render another image, then use an image processing program
to rotate the second image back 180degrees. When comparing, you should
see one pixel differences in polygon edges.

A true solution should be orientation independent. One such could be:
treat polygons as open sets and explicitly associate each edge with a
material. Similarly, treat edges as open sets and assign materials
vertices explicitly. That is, like the geometry is defined.

What do we gain by not going through this extra work? Clearly memory,
implementation clarity and a need to deal with degenerate geometric
entities.

--
Kalle Rutanen
http://kaba.hilvi.org
.

Relevant Pages

  • Oxford University landscape archaeologist Anthony Johnson: Stonehenge more geometry than astronomy
    ... The most complex geometrical achievement at Stonehenge is an 87-metre ... diameter circle of chalk-cut pits which mark the points of a 56-sided ... polygon, created immediately within themonument's perimeter earthwork. ... Stonehenge builders had geometry skills to rival Pythagoras ...
    (sci.archaeology)
  • Re: A question (JPG attached)
    ... you got me wrong, but thats my fault. ... High polygon looks good, ... > its hard to believe the method is globally broken for all geometry. ... >> or even a plain object. ...
    (microsoft.public.win32.programmer.directx.graphics)
  • Re: A shapes arbitrary contour?
    ... geometry shape for the square, rhombus, rectangle, triangle and then ... value of a custom property to a fixed value. ... For a regular polygon, the ... For a triangle, the first ...
    (microsoft.public.visio.developer)
  • Re: 2D geometry with M-file
    ... unclear to me and the cardioid example ... case you could specify a geometry as defined by a polygon. ... Download the code and you see inside the command-line code ...
    (comp.soft-sys.matlab)
  • Re: Icicles
    ... Maybe the Swarovski Avant Garde and Polygon Drop pendants in clear? ... Depends really on the other materials but if you are using silk flowers ...
    (rec.crafts.beads)