Currently, the Depth of Field (DOF) I am using takes a resulting depth buffer value and blurs accordingly. While this works OK for opaque geometry, I am having some issues with handling translucent geometry.
Eg. I either do not render into the z, in which case translucent geometry gets the DOF of what was behind it.
Or, I render the translucent geometry with z values (after the color buffer has been laid down) and then run into issues with looking through windows etc. (Distant mountains etc are not DOFed)
I have seen some people render a z that is based on how opaque the object is (eg. almost fully opaque, render the object z in the correct position, almost fully tranparent, render at a far distance).
I was wondering how other people are handling this issue. (or of they do DOF in a completely different way that I am missing something?)