1
$\begingroup$

I know that in Cycles Z pass is not anti-alisased while Mist pass is normalized anti-aliased. I need the real depth of each pixel (like the Z pass) but with anti-aliased. I don't know if getting that from the Z-pass (by making it anti-aliased) or from the Mist pass (making it not normalized) would be feasible and easier?

The real problem is when looking at the output Z pass, the pixels around thin objects like branches, grass stocks, are kinda sparse and missing. I doubt that could be because anti-aliasing is not on. I wouldn't want to post-process that with filtering, since they may create artifacts. So I am trying to get better anti-aliased depth image.

$\endgroup$

1 Answer 1

1
$\begingroup$

You can easily extend the mist pass again.

When measuring pixel values with ⎈ CtrlLMB LMB, be sure to set the display color color management to None otherwise your values will not be read linear.
enter image description here

  • Define an infinity limit, a value which is the maximum distance from the camera. Infinity will assume this value as well, objects beyond will clamp to this value. I used 1000. Set the mist depth to this value.
  • Set the mist start to 0.

Render a separate mist pass.

In Compositing:

  • Invert the Mist pass (1 - Mist).
  • Multiply it with the depth limit.

enter image description here

Remember to save as an .exr to retain values > 1.

$\endgroup$
2
  • $\begingroup$ In the figure, the depth multiplier, on the left (1000) and one on the right (100.00) are different. I thought they should be the same of 1000. Are they typos or done purposely? $\endgroup$
    – AugLe
    Commented May 15, 2018 at 8:43
  • $\begingroup$ It's a little different in my case: (1) My Mist pass only has "Start" and "Depth" (no "minimum" and "height") (not sure if it's the problem of version: I'm using Cycles with Blender 2.78a); (2) The results are mismatched if I do 1-Mist, but it's better to stay with just Mist*Depth, do you have any idea why? $\endgroup$
    – AugLe
    Commented May 15, 2018 at 9:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .