"Automagically" finding reversed slide scans and flipping them?

August 8, 2021 9:11 PM

Scanning a lot of slides, and I too-frequently put a slide in upside down so that when I view it on the computer everything is reversed. I'm looking for a tool to automatically find those and maybe correct the ones it detects.

I've searched for posts on slides and haven't seen this issue unless my search terms were bad, I don't think it has been asked before.

When I scan slides, it seems that I put them in upside down once in a while and the results are reversed. I'm scanning a lot of slides, four at a time on a flatbed scanner, trying to work in batches, so I don't detect the situation until it's loo late to find that one slide easily.

I know it can't be done for every slide, but it occurred to me that there might be tools that could find reversed letters like E, R, B, S, or Z (maybe others) to detect the situation and then possibly flip the digitized slide left to right to fix it. Does anyone know of programs or plug-ins (I have Photoshop Elements and Paintshop Pro among others) that might do this? While I'm asking - are there directories of plugins for these sorts of programs that would be a resource I could search before asking here?

posted by TimHare to Computers & Internet (2 answers total)

Do these slides all have at least some text? Is it contrasty? (IE: black text on a white background, or vice-versa)? If they do, or if they're high-contrast enough that you could get them there with level adjustment, you could try to run them through an Optical Character Recognition (OCR) process, and compare the output text to a dictionary list and see if most of the detected text is , then try flipping the image and repeat. If I were going to try this, I'd probably try Tesseract, a Python OCR library, and Pillow to manipulate the images (increasing contrast if needed, and flipping). The general algorithm would be "load image. OCR. compare each word in the OCR output to a dictionary file. Flip the image horizontally. Repeat. Which set of output text matches better against real dictionary words? That's probably the correct orientation.
posted by Alterscape at 11:24 PM on August 8, 2021 [1 favorite]

There simply nothing I am aware of that will perform the process you need. Software simply wouldn't "know" if an image was flipped.

An OCR app might be able to help you find the subset of flipped slides that happen to have identifiable lettering in them. In the case of lettering, there might also be edge cases where backwards letters might actually be the correct orientation. Lettering on a business window where your POV is looking out through the window, for instance.

If your scans include the cardboard frame of the slide, there might be hope on the OCR front, as the frames usually have some sort of text printed on them. That said, there's nothing out there, that I am aware of, that will automatically correct the flip, that I am aware of.

I'm afraid this is an instance where, to catch them all, you will have to go old-school and find/correct the scans yourself.
posted by Thorzdad at 4:35 AM on August 10, 2021