You are right. The only way to solve it completely in my opinion is ray tracing. However, according to OpenGL spec, the smooth implementation should calculate the coverage percentage to the pixels of the edge of the object, that means that if the situation you described occurs for two polygons of the same object, the effect you described won't happen. Only if two different objects edges falls in the same pixel and are not overlapping it will happen.
Thanks anyway, Guy. ________________________________ From: [email protected] [mailto:[email protected]] On Behalf Of Michael Platings Sent: Tuesday, March 09, 2010 7:01 PM To: OpenSceneGraph Users Subject: Re: [osg-users] automatic pixel transparency adjustment If "physically correct" is what you need then OpenGL probably isn't what you want. In general, OpenGL is suitable for rendering something that looks pretty good, FAST. I don't know of any way to get fragment coverage in a shader. However, even if you could that wouldn't be adequate: Imagine you have a white (255) square on a black (0) background. You size the square so that it precisely covers one pixel. When you render it, the pixel is coloured white (255). Now imagine you divide that white square in half. You would expect that rendering the two halves would also get you a value of 255, but... When you render the first half, the pixel is 255 * 0.5 + 0 * (1 - 0.5) = 127. When you render the second half the pixel is 255 * 0.5 + 127 * (1 - 0.5) = 191. NOT 255! Therefore, multiplying alpha by coverage wouldn't work unless you're only dealing with separate polygons, which isn't the case most of the time. HTH
_______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
