[Photoshop] How to Spot & Use the math difference between two images?

Justo

Hey guys, I'm having to deal with a particularly picky client that will post paintovers of the target goal he wants, expecting a 1:1 matchness, without mentioning what exactly his changes were when editing the image. 

Because communicating with him is difficult, I was wondering if there was a way I could overlay for myself in Photoshop his paintover with my version, set it to some blend mode and now exactly the HSL alterations that were made. I would think the Difference blend mode is the answer, but I'm finding it hard to apply that math to my version to obtain the same results. Any ideas?

Eric Chadwick

The way I have done this is to open their edited image, put the original as a new layer in a Group, select half, make that a layer mask for the group. Then add adjustment layers in the group, until it matches.

Eric Chadwick

If they're only making global adjustments, then that is how to reverse engineer.

If they did brush-based or masked edits, then making a flipbook style back-n-forth video might help.

You can also layout the two version side by side, then cross your eyes, overlaying them on top of each other. Your brain will pick out any variances automatically.

RN

I second the page-flipping method, you can do it with two layers on Photoshop, toggling the visibility of the top one. This is the easiest and best method.

 Just for completeness, you can also try loading the two images into Blender and using a Python script to create a third image with just the pixels that don't match. This will not work if the image was reencoded (JPEG for instance), since all pixels might look different.

Justo

The methods Eric points out sound perfectly workable, and is what I would normally do, yet my trust in the client is so low at the moment that even  a 1-to-5-RGB value difference might be reason enough for him to return my work. As you can imagine it's quite frustrating ha...

@RN That Blender method sounds very interesting. How could I achieve something like that?

RN

Justo said:

@RN That Blender method sounds very interesting. How could I achieve something like that?

You load the two images into Blender's Image Editor view, then open the Text Editor view, create a new text, paste this, change the image names in the script and run it:

https://pastebin.com/embed_iframe/Kp71USex

Blender is a convenient environment to run scripts that do something with image pixels. More information in here: https://polycount.com/discussion/205872/creating-images-with-python-in-blender

EDIT: Updated the code.

sprunghunt

RN said:

Justo said:

@RN That Blender method sounds very interesting. How could I achieve something like that?

You load the two images into Blender's Image Editor view, then open the Text Editor view, create a new text, paste this, change the image names in the script and run it:

https://pastebin.com/embed_iframe/Kp71USex

Blender is a convenient environment to run scripts that do something with image pixels. More information in here: https://polycount.com/discussion/205872/creating-images-with-python-in-blender

EDIT: Updated the code.

you can do the same thing in photoshop - put two layers on top of each other and set the blend mode of the top one to "difference"