OpenGL inverse depth calculation
Using gluUnproject is perfectly adequate for telling where a mouse click went in a 3D world, but when you have a large number of depth points that you want to transform (like when you are using the depth map of a mesh to perform visibility calculations) then it is rather slow and unwieldy.
So I decided to do the inverse depth calculations myself. However I couldn't find anywhere the maths to do the conversion from the value you get from glreadpixels (which is a float in the range 0 to 1) to an actual depth value (i.e. the z value in camera co-ordinates). After spending far too long just messing around with stuff and searching the web, I sat down and actually worked it out and it's this:
z = ( (2*zFar*zNear) / (zNear-zFar) ) / ( ((depth * 2)-1.0) - ((zFar+zNear) / (zFar-zNear)) )
where depth is the value you get from glreadpixels and z is the actual value you want.
Why can't they just put that in the docs somewhere?
Joe
Re: OpenGL inverse depth calculation
Hi...
Finally I find something in relation to this.
I have tried your equation and it works fine if the zFar/zNear relation is not big.
I tried it with zNear = 2, zFar = 4 and it seems to work ok, but if I begin moving the zFar plane backwards, it begins adding error to the real z.
Have you made any improvements to this equation?
Cheers,
Victor
Re: OpenGL inverse depth calculation
Hi Victor,
I haven't made any improvements to it. Like I said, I simply reversed the maths in the documentation and this is what came out. Maybe there are some extra things I missed out that don't affect the ranges I am working in.
Try looking at the code for gluUnProject if this isn't working for you.
Joe
Post new comment