1

I've got a scanned picture of a sticker album which contains an translucent room where the sticker should be sticked. I've tried (using the Equalize, Curves, Levels, Vibrance and Contrast Tools from Photoshop) to get the translucent room as close as possible to the real sticker, but I feel that I still can do something to improve it. Could someone give me, kindly, some hint about this situation?

Example of the problem mentioned in the question.

P.S.: In case of doubts, like the user Scott appointed me, I don't have the image B. That which one that I used like example it was found on the internet. They are 100 stickers and I found some of them on a website, but I'd like to "create" the others ones throught the faded images that I have in the empty album (in this case, those images A).

12
  • 3
    You cannot accurately recreate 2 from 1 because you have only 25% of the information necessary to do so. Original-size images may help, but it's still an unlikely rescue. – Tetsujin Feb 14 at 19:46
  • 3
    Kind of begs the question... if you have B why do you need to bother with A? – Scott Feb 14 at 19:54
  • 2
    The use of the word "opaque" confuses me. It means the opposite of "transparent". Perhaps "faded" would be a better word? – Wolff Feb 14 at 21:41
  • 1
    Translucent perhaps? :) – Scott Feb 14 at 23:30
  • 1
    I agree with @Tetsujin here. Perhaps if you posted a full size scan at high resolution (or a crop of it) we could give it a go, but even then I doubt there is enough information. Not only would the scan have halftone dots but you also have sort of "pixelated" color information. It's impossible in the same way that you can't scale an image down and then scale it up again without loosing quality. – Wolff Feb 15 at 0:32
2

I'm able to get a result that's a tiny bit better than yours using mainly Camera Raw Filter.

Method for restoring original colors

This is the starting point (scaled down to 25% to be able to post it):

  • Start by turning your image into a Smart Object by right-clicking the background layer and choosing Convert to Smart Object. That way the filters are dynamic and can be fine-tuned later.

  • Use Filter > Camera Raw Filter with the following settings under the Basic tab to restore the colors:

    Under the Detail tab you can use the following settings to remove some of the halftone noise:

  • To remove more noise, use Filter > Noise > Dust & Scratches.

  • Smoothen even more with Filter > Blur > Gaussian Blur.

  • Finally, try to recreate some of the lost sharpness with Filter > Sharpen > Unsharp Mask.

I think it's a decent result, but probably only fit for viewing how the stickers looked. Not really suitable for reproduction.

To scale the image down can further camouflage some of the problems (and I can't post the full size result here). Here is the result scaled down to 25% using Bicubic Smoother interpolation:

Why it isn't possible to restore the original colors

It looks like the images have been faded to about 20% of their original opacity or less.

The pixels of the original images could potentially have RGB values ranging from 0 to 255. If they were faded to 12.5% they would be compressed into a range going from 223 to 255. So a lot of the data is lost. It can be compared to taking a 256 pixels tall image, scale it down to 32 pixel and then up to 256 pixels again:

The color information has become "pixelated"!

And then you have the additional challenge that you don't even have access to the original pixels. You only have a scan of the printed halftoned images. This just adds to the problem. In a halftone screen the different tints are expressed by dots of different size. These very light images consist of very small dots with very subtle variations in size. Furthermore the scan also contains smudge and paper texture. Almost impossible to recreate the original tones from that.

4
  • You can vur and sharpen too. – joojaa Feb 15 at 22:06
  • Instead of Gaussian Blur and Unsharp Mask? – Wolff Feb 15 at 22:10
  • 1
    Smart sharpen is better it does a statistical inverse gaussian blur (or lens blur). But since the data is distributed it should just kill the moire pattern as you can afford to blur more. One could also try minimum before blurring. – joojaa Feb 15 at 22:49
  • 1
    Thank you so much for replying, @Wolff. It helped me a lot! That's a great result! And besides I could learn a little bit more about some tools that I didn't know. Thank you for helping and for so great enlightening! God bless you! – winiercape Feb 17 at 0:04
1

Just as a basic starting point...

I'd take the scan and duplicate the layer and set the duplicate's blending mode to Multiply. Then repeatedly duplicate the duplicate. This will darken everything with each new duplicate. Repeat until you feel you've reached a general medium.

Then add a levels adjustment layer to tweak some values, decreasing mid-tones a bit, bringing up shadows, dropping highlights.

enter image description here

From here, I'd focus on individual pieces.


Why not simply use Levels/Curves adjustment rather than duplicating a multiply layer?

While a simple Levels adjustment may get you close to the results the multiply layers get you, there's a difference.

When multiply layers are used, they duplicate layers react based upon underlying pixel values. Whereas a levels adjustment merely alters existing pixel values. The build up of values is more subtle and tonal with the multiply layers. Rather than sweeping adjustments to all pixels, only pixels which can be darkened, based upon that pixels value, are darkened.

This is a common technique for darkening a faded, or partially transparent image in order to get closer to 100% values for the pixel.


Even given this... I'll be honest, in my opinion, removing the inherent grain found in the images will be exceptionally difficult.

I do not know that a high quality reproduction would be possible without painstakingly painting in some colors manually. Similar to fixing that scar down the middle of the center column - it's going to require dedicated manual attention.

1
  • Thank you for your answer, Scott! I'll try your hints too and I believe they will help me a lot like Wolff's answer. I appreciate your support, man! God bless you! – winiercape Feb 17 at 0:07

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.