View Full Version : Digital Confusion is confusing Antialiasing

07-01-2003, 11:55 PM
I have this [very] simple scene set up with depth of field focused on the mug using Digital Confusion (autofocused on a null object placed at the lip of the mug), f-stop of 1.0, otherwise default. Without antialiasing, Digital Confusion does its thing nicely, but with it everything not an edge is tinted with the color of the backdrop. Anybody know why?

07-02-2003, 07:15 AM
I've given up on Digtial Confusion. It added a noticeable halo around the focused object and gave it a degree of transparency. Your samples demonstrate its problems even more dramatically.

I haven't noticed any transparency with the DOF Blur plugin, but it can add a halo.

The built-in DOF seems to be the most reliable. When I asked about problems with Digital Confusion, a couple of people recommended X-Dof (http://evasion3d.com/xdof_intro.html).

07-02-2003, 10:16 AM
Thanks for the feedback. I've not used LW in over 2 months, but I don't recall noticing any issues like this with Digital Confusion then. Maybe that was before the 7.5 upgrade; I can't recall. I'll check out X-Dof.

07-02-2003, 06:34 PM
Do you have "render hidden geometry" turn on? The halo comes from blurring when there is nothing rendered behind the mug.

I don't like DC either btw

07-02-2003, 09:14 PM
I tried with and without Render Hidden Geometry and got the same result. It's not crucial that I use Digital Confusion. I'm just learning the software now. Thanks for your input.

07-03-2003, 05:34 AM
I would strongly recomen 'defocus dei' from Blackfeet.co (I think), which, while not using a hidden geomerty render, gives you enormous control without having to do hundreds of test renders. It's plugin saves the z buffer in it's own format, and you then open the image in the defocusdei interface and play with the settings 'till you get what you want. You can keyframe all the settings for animations.
I often run it as a default with all my animations, it takes very little extra render time, and (as long as you dont run out of frame-store space) it's application can frequently save a lot of re-endering.

Of course it doesnt work with volumetrics, in fact it makes a right mess of light beams - because the slightest invisible edge of the light beam replaces the z buffer number as the distance to that pixel.

No-ones perfect after all.

07-03-2003, 06:07 AM
I had the same problem. I re-rendered the scene with enhanced AA and it worked fine.

Hope that helps:)

07-03-2003, 10:49 AM
I had the same problem. I re-rendered the scene with enhanced AA and it worked fine.
I just tried that and the image was hugely improved. Now that I've achieved that, however, I see other failing details: black ghosting around the bright edge of the mug and lines between blurs on the desk plane.

The built-in DOF seems to be the most reliable
Quite. I'm nearly given up on Digital Confusion.

Most of the DOF plugins I see provide different qualities of blur, but few of them keep the brightness of lights (downfall of post-processing, I guess?). That's one of the key elements that makes DOF look photorealistic instead of just blurring layers in Photoshop. X-DOF seems to do that well, can anybody recommend some others for comparison?

This image is a similar shot with Enhanced Extreme antialiasing...oh well.

07-03-2003, 11:47 AM
i had same DOF issues. so i put this together.

loads a scene with a camera's DOF attached to a null objects.
uses LWs internal DOF.

07-04-2003, 05:40 AM
Hi Flash,

Have you tried turning off 'Adaptive Sampling' in the camera object properties? I'm sure I read somewhere it causes DC to become even more confused(ing)!!

See ya

07-04-2003, 09:42 PM
Granted, it's still not perfect, but it looks like that's the best DOF with Digital Confusion is going to get.

You can see the difference in render time made by turning off Adaptive Sampling, but the image quality improves drastically (still some ghosting around the object in focus, though).

07-04-2003, 09:49 PM
personnally I wouldn't be happy with that result - has anyone mentioned rendering a z-depth buffer and using it for dof blur in post? I'm sure it's what the major studios do