Retrieve normal maps difference
polycounter
This might be rather simple but I can't figure this out. I have 2 normal maps where the only difference is that one has more cracks. Now I want to somehow make a normal map out of these which is only the cracks, the cracks on flat.
The first thing that pops to mind is to just use difference in the Photoshop layer options but that doesn't do the trick. I'm kinda lost here, maybe it is not even possible after all?
The first thing that pops to mind is to just use difference in the Photoshop layer options but that doesn't do the trick. I'm kinda lost here, maybe it is not even possible after all?
Replies
Copy and paste your red and green channels from your two normal maps into different layers. Blend the two red channels using Subtract, and put them in a group. you'll notice you only get the + side of your red channel, that's why you now have to duplicate the group and invert the Red channels, this will give you the negative side of the red channel. I tried combining the channels with a screen @ 50% opacity, and for some reason instead of middle grey being 128 it was @ 108, I dunno, repeat for the Green channel.
I got this by taking one of the normal maps, inverting the red and green channels with two filter layers set to Copy Red blending mode and Copy Green blending mode respectively, then putting the other normal map on top of it in Combine Normal Maps blending mode. You can take a look at the .kra here: https://dl.dropboxusercontent.com/u/15437561/krita_normalmap_difference.kra
Because it uses reoriented normal map blending to combine the normal maps, this method gives superior results to the Photoshop way. If you have another utility lying around somewhere that also does reoriented blending (surely there's a PS plugin for it around somewhere) you could do it with that, although it would be a little more destructive.
Edit: if it's not swizzled the right way and looks wrong and stuff you can switch them and the result is a different.
did the same thing but in Adobe AfterFXImportant! - Project MUST be in 32 bit mode
so...
first we substract version without cracks from version with cracks (layers 2 and 3)
and then add top layer which is "flat" normal color
you can use similar workflow to extract bounced lighting, caustics and shadows from 3d renders
32 bit mode is important as it will allow you to use values below 0 (negative values)
by substracting "non cracked" texture from "cracked" you get differences in images in values from -1 to +1
and then on top we ADD the bland flat normal texture so you get real differences between cracked and non cracked texture (isolated cracks) applied to the "flat" tetxure
@Scruples - you didnt get expected value from Screen as it is using different blending formula:
invert ((inverted textureA) * (inverted textureB))
instead of simple A+B
Another approach, if you're certain that every pixel in the textures except the cracks match, is to write a simple GIMP Python-Fu script to compare each pixel in two layers and make a third layer with the pixels that don't match (based on one of the layers, like the topmost one for example).