Photos with color deterioration

[Greg325941064b6f]

I have a number of old scanned photos where the color has deteriorated.  I've tried correcting the skin tone with Levels and Curves but I cannot get the skin tone to look natural.  Hoping fellow users may have some suggestions for me. 

[kglad]

what app?

[Greg325941064b6f]

Sorry....Photoshop desktop app.

[kglad]

in the future, to find the best place to post your message, use the list here, https://community.adobe.com/

p.s. i don't think the adobe website, and forums in particular, are easy to navigate, so don't spend a lot of time searching that forum list. do your best and we'll move the post (like this one has already been moved) if it helps you get responses.

<"moved from cc desktop">

[Greg325941064b6f]

Many thanks

[Kenneth Kawamoto]

This is what Photoshop Neural Filter thinks... 🤔

[Greg325941064b6f]

That is certainly better skin tone than I've been getting.  I've been using the Neural Filter / Photo Restoration.  I'll continue to tweak.  Thanks

[Kenneth Kawamoto]

I actually used the Colorization filter (with mask), not the Restoration filter - on top of Camera Raw filter to remove the cyan cast first

[Greg325941064b6f]

Awesome, love learning new stuff!  Many thanks!   Here is my tweaked version.

[Conrad_C]

Wow, this shows some of the limits of AI. It looks like the AI doesn’t recognize the uneven color shifts, and has not been trained to understand that the blue cap and gown should be the same blue throughout. It thinks the magenta shift is original. It has obviously been trained on skin tones, though.

[D Fosse]

I only had time for one -

This is more difficult than it looks. The problem is, of course, inks/dyes fading - but not uniformly. Each color is faded to varying degrees and in various areas of the image, and while one is faded in the highlights only, another is faded further down.

In short, the individual curves are all messed up.

Here's my layers panel. All the masking is done with very soft brushes at low opacity.

[Greg325941064b6f]

Something about the background in this photo seems to cause various problems as well.  If this is the original background, (and not something deteriorated), it's the most awful in history.  After I replaced it with something more neutral then things started working better.   This image is simply using Colorization, but I'll try it with your suggestions as well.  Thanks!

[Trevor.Dennis]

That's pretty good.

[Kenneth Kawamoto]

Again I let Photoshop AI to colour this one. I think it's impossible to restore colour with curves

[didiermazier]

Alternatively to smart filters you could simply set the white and black points (with curves or level). For most of cases it gives fast and good results

[Stephen_A_Marsh]

Alternatively to smart filters you could simply set the white and black points (with curves or level). For most of cases it gives fast and good results

By @didiermazier

This should always be step #1 if such a point can be identified, just as setting the correct white balance is in a raw capture. The shirt collar would appear to be such a candidate in the first example. The white point can always be set in colour blend mode so that luminosity can be adjusted separately.

[didiermazier]

Also if you work for print you should set WP to C3 Y2M2 in order not to avoid what the guy at the printing call the "hole in the paper" except for specular white (i.e. strong reflection on metal). It is neutral but with just 1% more C for contrast.

The black point should not exceed 250 % so K85Y50M50C60 should set up a rich black. They call excessive black "buttering the plates"

Again all this for printing in Europe, don't know in the USA

😉

[D Fosse]

This should always be step #1 if such a point can be identified, just as setting the correct white balance is in a raw capture.

By @Stephen_A_Marsh

Yes, but what actually made this a lot more difficult is that the scan is clipped in the blacks. That not only removes the native black point, but also makes it impossible to determine any color channel relationships at all in the shadow region.

[D Fosse]

OK, I'll come clean. As some of you may have noticed, I think AI is the devil's invention and what will eventually eat our whole civilization up from within. Not an explosion, but a silent implosion.

And I've said so for a long time. But - just to be positive, I have also said that it has good uses, good purposes. I've just been struggling to figure out what they could be, aside from curing cancer.

But this - this might just be one! Restoring old and half-destroyed photographs. That's a worthwhile purpose for AI. It's just too difficult, often impossible, to do manually. When disaster hits and people's homes are destroyed - what's the first thing they rescue? Right, their old photos. Memories are important. Restoring old photos is important work.

[Greg325941064b6f]

Agreed.  My mother and aunt both wept when they saw this photo I restored of my grandfather.

[D Fosse]

Yeah, that's a splendid example. Luckily here the important parts weren't too damaged - but still - gruelling work that AI does better and much quicker.

[Conrad_C]

I have a number of old scanned photos where the color has deteriorated.  I've tried correcting the skin tone with Levels and Curves but I cannot get the skin tone to look natural.

By @Greg325941064b6f

I’m in the middle of scanning old color negatives, and these problems look so familiar. And they are not easy. Most of the problems have to do with color dye layers fading at different rates, but as in your examples the problems can also vary spatially because other color shifting seems to be starting in from the edges. That usually means two separate corrections to address each of those problems.

Levels, Curves, and White Balance corrections often fail with faded film images because (I think) they’re not really set up to correct the way the cyan, magenta, or yellow dye layers fade. I have had some success by making the Channel Mixer the first step, because Channel Mixer lets you boost faded color channels using info from other color channels. And, it shifts all values of a color by the same amount, unlike how Levels and Curves usually work. Selective Color is also a good first step, especially if you think it would be better to target CMY while keeping the image in RGB.

Here’s what I got. In the first image, the heavy lifting is done by a Selective Color adjustment layer to address the overall color cast. But that method wasn’t working for the different fading in the center and face, so I gave up and used the method of reversing a color cast by sampling it from the very convenient white collar, and applying that through a Solid Fill color layer set to the Divide blending mode. (I learned that as an Adobe After Effects tip, but it also works great in Photoshop.) Then, using a layer mask, I painted in that correction where needed. The Vibrance and Curves layers just do some final minor tuning.

In the second image, the primary correction is using a Channel Mixer adjustment layer, but again I needed another layer group just for the face. That layer group uses Channel Mixer and Vibrance adjustment layers, limited to the face by a mask attached to the layer group. That mask was initially built using a selection from Select > Color Range, but I refined that by painting in it. I am still not happy with the skin tones, but it got most of the way there. Again, the Vibrance layer is just a minor tweak. So both images are mostly corrected by one or two Channel Mixer or Selective Color adjustment layers.

[Trevor.Dennis]

Your last image is about right I would say Conrad.  IME red haired people tend to have pale skin.  Some of the worst subjects I have had to try and get a decent picture of were red haired smokers.