[Bf-committers] DOF camera parameters.

Alexander Ewering bf-committers@blender.org
Thu, 10 Jun 2004 13:25:39 +0200 (CEST)

On Thu, 10 Jun 2004, Jonathan Merritt wrote:

> Hi Alexander,
> Indeed, both circumscribing the circle-of-confusion and z-blur are both 
> hacks...
> In the z-blur method, hiding has already been done prior to calculation 
> of blur.  The advantage that the CoC method has is that hiding is *not* 
> done prior to calculation of the blur.  Hence, objects with a large 
> circle-of-confusion can blur out from "behind" objects that would 
> otherwise hide them (which happens in reality due to the fact that the 
> lens and aperture of a real camera have a non-infinitesimal width).  
> However, even the CoC method (as described) is still a hack, since the 
> inside of the circle of confusion is not sampled.

The biggest problem the circle-of-confusion method has, in my opinion,
is that it is awfully slow.

> Shouldn't z-blur be possible as a filtering process (provided you can 
> access depth information for each pixel)?

Yes, that's how it works. It uses the zbuffer supplied by Blender
(most of the blurring code is taken from strubi's zblur plugin)

| alexander ewering               instinctive new media
| ae[@]instinctive[.]de   http://www[.]instinctive[.]de