would anyone care to offer advice on the best way to combine 2D, vector driven motion blur with 2D DOF during compositing? Of course, I'm dealing with 3D rendered output.
My main concern is the ordering of the blurring effects, so that the most accurate final output can be obtained.
If motion blur is applied to the composite before DOF, the same moblur must be applied to the depth channel. In this case should my depth channel be treated in any way? I'm accustomed to rendering a no-aa depth channel, for more accurate depth. If the depth channel is blurred, won't the depth values for moving objects be quite incorrect? Would applying a rank filter, for example, be helpful to maintain correct depth values?
What if DOF is applied before moblur? Surely, applyig DOF to the motion vectors will yield bizarro, outright wrong results from the motion blur? Or am I mistaken?
Based on my above stated reasoning, in the past I've applied motion blur, followed by depth of field to my composites. My results have been quite satisfactory, but I've never been sure if I'm doing this right and I've always felt that some amount of repair was required on the depth channels.
Would anyone care to share how they deal with 2D motion and DOF blurs? I'd love to hear what more experienced compositors do in this case.
P.S. I'm generally working with node based compositors e.g. XSI, Nuke, but I assume the same principles would apply to something like After Effects.