Forum: Carrara


Subject: Depth of field wrecks anti-aliasing

AndyCLon opened this issue on Nov 11, 2007 · 4 posts


AndyCLon posted Sun, 11 November 2007 at 4:13 PM

I'd been looking at some techniques for [ turning real photos into models](http://fleacircusdirector.blogspot.com/2007/10/more-depth-of-field.html) by adding faked depth of field and thought that this would be an excellent idea to add to my flea circus.

However, when I turned on the depth of field, I got horrible jaggy edges, see attached example.

Is there something I can do to eliminate or avoid this problem, this will eventually be a 5 minute movie with 25 frames per minute so I don't want a manual fix.


Miss Nancy posted Sun, 11 November 2007 at 4:35 PM

it looks promising, but the image is too small on my monitor to discern any jaggies. my method would be to import the layers into a video editor and blur 'em there.



MarkBremmer posted Sun, 11 November 2007 at 7:18 PM

Hi Andy, Ray traced DOF will fix the antialiasing issue. Howver, it takes longer to render. The second option is to use the regular antialiasing but render the file larger and then reduce the image size on final movie export. This takes longer to render too but usually not as long as the Ray Traced DOF. A third option is to render the DOF g-buffer and then use the g-buffer channel to drive a focus parameter in AE, FCP, Motion or Shake. This will create a slightly soft look becasue there will be a little blur creep at edges where the DOF changes significantly. The last option is to render scene layers as separate movies and then composite them and do the DOF in your video editing application. Mark






AndyCLon posted Mon, 12 November 2007 at 1:11 AM

Thanks, these gave me some ideas. The difference between the quick and full raytrace is 1s to 3minutes and does fix the problem with a significant hit.

I've only got Adobe Premiere Elements and it's not capable of the compositing effects described as far as I am aware.

Thanks to the comments I did have an alternative idea. If I render the 2 background objects with Depth of Field and then use the resulting image as a background, rendering with those background objects invisible then I should be able to get the best of both worlds.