2
\$\begingroup\$

I digitized a double exposure film negative taken 50 years ago. (It is a .tiff file) Is it possible to remove the superimposed images using Phtotoshop or Python? enter image description here

enter image description hereenter image description here

\$\endgroup\$
2
  • 2
    \$\begingroup\$ Do I see correctly that you have on the left the double exposure, and on the right an image of one of the single exposures? \$\endgroup\$ Commented Nov 3, 2023 at 13:10
  • \$\begingroup\$ Yes, that's right. \$\endgroup\$
    – moruna
    Commented Nov 4, 2023 at 3:05

3 Answers 3

4
\$\begingroup\$

Possible, but difficult, and likely only bringing imperfect results.

If the second image and one of the double exposed images were perfectly identical, being taken in the exact same place, angle, and with exact camera settings and film, this would be very easy, just subtracting the images, because all a double exposure is, is adding two exposures. In GIMP this would be done by placing the single exposure on the layer above the double exposure, and changing its mode to "Subtract". In the case this doesn't produce sufficient results, it may be the case that curve adjustments may need to be made to correct for non linearity of film.

However, they are not identical, that we can see even with these lower resolution scales. The image without double exposure would need to be manipulated, corrected, scaled, and changed, in order to match the double exposure. Once they have been made sufficiently identical, the process described above for identical images would be followed. The exact manipulations would be specific to the exact image. Even after that though, the images are not perfectly aligned, so some areas of the double exposure are impossible to correct for, as the second image simply does not contain the information from those areas.

\$\endgroup\$
5
  • \$\begingroup\$ Could you perhaps go into more detail on the case that the images would be perfectly identical? How can one subtract the layer and be done with it? What steps does one need to roughly follow (in PS, GIMP, whatever program you choose)? \$\endgroup\$ Commented Nov 5, 2023 at 9:13
  • \$\begingroup\$ @SaaruLindestøkke Is it better now? \$\endgroup\$
    – Topcode
    Commented Nov 5, 2023 at 17:05
  • \$\begingroup\$ I just tried it out with some dummy images and not sure I got the right result. I started with this image, let's assume the serrated lines are the double exposure we want to get rid of. Here are the serrated lines separately, note the background is transparent. Now when I put the separate lines above the image and set the mode to "subtract" I get this. The serrated line is still there, but now simply in shades of grey. What am I missing? \$\endgroup\$ Commented Nov 6, 2023 at 9:54
  • 1
    \$\begingroup\$ @SaaruLindestøkke the issue appears to be that the lines have an alpha channel and are partially transparent, which is messing it up, try using an image without an alpha channel. \$\endgroup\$
    – Topcode
    Commented Nov 6, 2023 at 17:12
  • \$\begingroup\$ Thanks, that seemed to be the issue, I know get a neat, full black mask of the serrated lines that can be used in subsequent filling steps. \$\endgroup\$ Commented Nov 6, 2023 at 18:32
2
\$\begingroup\$

I would say it's nearly impossible. If the image on the right would be an exact copy of the image that overlays the photo of the person I imagined you could do something like:

  • preprocess image on the right to mostly show outlines of the objects
  • use it as a mask on the photo of the person to delete the pixels that match the objects
  • use (generative AI) fill to restore the photo of the person

I tried it out with the (miniature) photos you posted for about 15 minutes without any noteworthy results. As I am not a Photoshop master, nor do I have the higher resolution images I can imagine that my experience is not final, so feel free to try out the steps above yourself.

\$\endgroup\$
-1
\$\begingroup\$

If your Photoshop skills are up to the task try making the image to the right as a channel to make a selection on the portrait image. Modify the selection with feathering or similar to soften the edge and use the CONTENT AWARE FILL tool. I also hear that the new PSD 25.0 Has a 'generative fill' but have not used it personally. Increasing the pixel size of your image will help, if you can rescan the original.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Could you give an example of using the techniques you describe here? \$\endgroup\$
    – Philip Kendall
    Commented Nov 5, 2023 at 14:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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