1
$\begingroup$

When I invert the colors in Blender, I don't get the same result at all as Photoshop. The middle-grey background should not change, but it gets brighter in Blender while it stay the same value in Photoshop.

$\endgroup$
7
$\begingroup$

Blender will generally linearize colors when importing 8bit images. Photoshop does not. Normally, Blender's behavior is correct but in this case you are encoding data here, not colors. You can inform Blender of this and disable the de-gamma step by setting your input color space to "linear" or "raw" instead of "sRGB".

$\endgroup$
  • $\begingroup$ Thank you for you answer. I had already tried the linear and raw mode, the background grey doesn't change when i invert the image, but the source image appears brighter and the result is brighter than Photoshop. $\endgroup$ – Danyl Bekhoucha May 30 '16 at 17:09
  • $\begingroup$ Blender will also convert back to non-linear when saving, this seems to be the other issue. When I set color management to "none" AND the input space to "linear", the result is the same as Photoshop. Not sure if there's another way to disable gamma correction of the final output apart from disabling CM or saving as OpenEXR. I thought "save as render" did this, but it doesn't seem to work here. $\endgroup$ – JtheNinja May 30 '16 at 17:21
  • $\begingroup$ It works but it becomes hard to manage the other images and colors. For example an image attached to a multiply node with a completely white color becomes darker. $\endgroup$ – Danyl Bekhoucha May 30 '16 at 18:04
  • 1
    $\begingroup$ Yeah, globally disabling CM is a very poor solution, it should be bypassed specifically for saving the cavity map. Just not sure how to do that. $\endgroup$ – JtheNinja May 30 '16 at 19:12
  • 2
    $\begingroup$ That also won't quite nail it. The problem is that an inversion is on display referred encoded values, and as a result, a 2.2 power curve won't invert entirely 1:1 with Photoshop, and it certainly won't work at all on scene referred data coming from Cycles. "Inversion" of a colour has no real-world equivalent, and is rather a hack to achieve something else. For alpha, which is always a bounded 0..1 linear ratio, 1.0- value will work correctly. $\endgroup$ – troy_s Jun 5 '16 at 0:28
1
$\begingroup$

Here is the node setup to invert the colors properly:

enter image description here

$\endgroup$
  • $\begingroup$ Using this, I do not get the same depth map back. Why is that? $\endgroup$ – Amir Mar 10 '18 at 7:23
-1
$\begingroup$

Here's a visualization of how you can change the color space to "Raw" so that you can do inversion properly:

enter image description here

Before inversion:

enter image description here

After inversion:

enter image description here

And for depth maps (connect the Z-pass to a Normalize node first)

Before inversion:

enter image description here

After inversion:

enter image description here

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.