– might be interesting, for the record, to describe what I have tried with my scan setup:
- White LED, photographic camera(s), single exposure
- White LED, Raspberry Pi v1 camera, single exposure, automatic and manual settings
- Separate R/G/B-LEDs, tunable intensity, Raspberry Pi v1 camera
- Separate R/G/B-LEDs setup, but with different main wavelengths, Raspberry Pi v1 and v2 camera with lens shading correction
- Started to research how to handle the massive dynamic range color-reversal film exhibits:
5.1. taking between 2-20 exposure of a single frame, creating a real HDR out of this. Problem at that time: there were no real HDR-standards available for post-production and distribution, so this approach was dropped at that time. Might revisit this one
5.2. taking raw images (10-12 bit) - problem here: post-production was too manual and 12 bits is really the absolute minimum you need to capture for color-reversal film. No margin of exposure error.
5.3. creating a fake “HDR” by Merten’s exposure fusion - this was the route chosen, as this algorithm maps the dynamic range mostly automatically into the 8bit dynamic range normal displays and distribution channels can handle. - 'cause lens shading is not able to to compensate the color-cast of v1/v2 cameras (v2 much worse than the v1 camera) paired with a Schneider Componon-S, switched to a cheap USB3-machine vision camera
- re-switched back to Raspberry Pi cameras, this time the new HQ camera - no noticeable lens shading issue (there are, but they are not really noticeable)
- Switched from the mode “multiple exposure times for a single frame/fixed illumination level” to the mode “fixed exposure time/multiple illumination levels for a single frame”. This speeds up the image acquisition considerably.
A few further remarks on the last step in this list: you will need a very good DAC for this, at least 12 bit for fast switching of the LED-source to different illumination levels. You still will be limited by the available dynamic range of the LEDs - they need a minimum current to turn on.
Anyway, there are appropriate discussions and designs about this topic to be found here on the forum. Be prepared to notice color shifts between the different illumination settings: every LED shifts color when driven with varying currents (causing slight temperature shifts which in turn shift the primary wavelength the LED is emitting). So your color balance will be slightly off between the exposures you are trying to combine into a single image. This color shift will also be a function of the mean temperature of the LEDs, so it will be different if you grab the exposures from dark to bright vs capturing the reverse sequence, bright to dark. This might lead to a magenta cast for the highlights, a green cast for the mid-tones and a yellow cast in the shadows.
So be prepared to walk through some extensive tuning, introducing possibly different color-balances for each exposure level. Given, in principle you should be able to counteract these color shifts in post-production, but I have not tried this route so far. In my setup, I take five different exposures, each with an individual color balance setting.