Naked Science Forum
Non Life Sciences => Technology => Topic started by: Pseudoscience-is-malarkey on 14/10/2023 15:24:34
-
I read an article about a man selling his idea that there is a "DNA" in monochrome photos and video stills that reveal the actual color of objects and such by the shade of grey they project. My hang-up on this is he doesn't seem to acknowledge one important variable: shadows. Shadows of course affect the hue, which must sometimes fool AI in the final product. How can AI compesate them?
-
A lot of monochrome photography uses color filters to enhance contrast (blue filter enhances the lips but can make a slightly stressed eye look really diseased), rebalance the overall scene to compensate for the film response ("pan" or "ortho" being familiar film stocks with quite different spectral sensitivities) or produce a "true night" effect (mimicking the spectral response of the human eye at low light levels) so the final image can't map back to the original color palette.
You can use facial recognition to bring a bit of life to an old photo but it has to be exceedingly clever not to make everyone the same shade of pink!
Testes, or bovine feces?
-
You're right about shadows being a tricky part for AI when it comes to interpreting colors in monochrome images. The AI has to be trained to distinguish between true color information and the effects of lighting and shadows. One way it can compensate is by learning from a vast number of images where the color values are already known. This way, it can start to recognize patterns and make educated guesses about the colors in monochrome photos, even with shadows present. It's all about the AI's training and the algorithms used.