Anteru
05-27-2003, 09:23 AM
I am currently implenting DoF in the free OS renderer Aqsis, and I
encountered a quite big problem.
The CircleOfConfusion doesn't seem to be right.
I'm using the one from the RiSpec, so I assume that I pass some wrong
vars to it or that I use it's result wrong. I compute the average
depth of the MPG ( in "camera" space), and pass it through to the CoC.
Here is my first question: Should it be the depth from the camera, or
should it be measured from a different point? Then I get a CoC value.
Now should I scale it by a value or not? I mean, the CoC could become
an ellipse on the image plane. It has been computed in camera space,
now I need to transform it onto the image plane. This should be easily
done by transforming the vector from "camera" to "screen" space, but
for some reasons this doesn't work as expected. I get either too much
or not enough blur.
Thnx for your help!
MGC
encountered a quite big problem.
The CircleOfConfusion doesn't seem to be right.
I'm using the one from the RiSpec, so I assume that I pass some wrong
vars to it or that I use it's result wrong. I compute the average
depth of the MPG ( in "camera" space), and pass it through to the CoC.
Here is my first question: Should it be the depth from the camera, or
should it be measured from a different point? Then I get a CoC value.
Now should I scale it by a value or not? I mean, the CoC could become
an ellipse on the image plane. It has been computed in camera space,
now I need to transform it onto the image plane. This should be easily
done by transforming the vector from "camera" to "screen" space, but
for some reasons this doesn't work as expected. I get either too much
or not enough blur.
Thnx for your help!
MGC