When using POV-Ray to render sprites for a game and using AA, the edges
of the sprite are AAd with the background. So I have been thinking of
modifying the way AA works as follows.

Count the number of supersamples which hit objects and the number that
hit nothing.
Add up the colour values of the 'hit' rays only and divide by the number
of 'hit' rays.

Now the objects should be AAd but still have sharp edges and not contain
any of the background colour.

Another option would be to AA only the transmit value with the
background and thus get soft (transparent)
edges so rendered objects could be pasted over other images in a paint
program. I was trying to add POV rendered buildings to a photo and the
object edges were driving me crazy.

Anyway, does this sound like a sensible idea? Has anyone done it before?
Anyone see any problems?

Thanks, PoD.

Post a reply to this message