High-Resolution Film Scanning with a Flat Bed
Status: Hypothetical & Preliminary Testing
Scanning film with a flatbed doesn't result in large detailed sharp images unless it is large format, and medium format for some good flatbeds that achieve 2000 dpi.
Any image sensor is capable of resolving 100% of its resolution, whether it be a compact digital camera, flatbed scanner, or otherwise, but they do not, in the case of compact cameras their resolutions are so high these days, that you would need a really expensive prime lens resolving near the diffraction limit at f/2.8, instead they have cheap and often slow zooms who's diffraction limit is already below the sensor's resolution.
In any case, on to scanners, a scanner's resolution isn't beyond resolving by a cheap but good film prime lens for example, 4800 dpi resolution is 5.3 microns and as such has a diffraction limit of f/9.6, it's not too hard to get a prime to be damn sharp at f/5.6 to stay safely away from the diffraction limit, but then we dont need to resolve 4800 dpi out of film (unless you have a high resolution film and the lens you shot the original image with was under the diffraction limit as well.. the same f/9.6 limit will also apply here if you are looking for 4800 dpi of detail).
Now a lot of scanners use an array of pinhole lenses over the sensor - this is one reason why they are cheap and have such a shockingly lower resolution than the sensor is capable of.
There are a couple of ways we could over this, the most complex way would be the route of the typical scanner camera project - disable the LEDs, and pull out the pinhole lens array, mount a lens and make it light tight, then the scanner can pick up whatever resolution the lens can feed it, thus you could backlight some film and project it directly onto the sensor.
There are a couple of problems with this method, one is that you typically need the LEDs and pinhole lens array and platen glass left on for the scanner to pickup and calibrate on the grey strip under the platen glass off to one side, usually it needs all the components there to do this, but you can trick it, though your images will look poor (if any because of it), have a look at any diy scanner camera project, the quality is discouraging (apart from the professionally built scanner cameras like BetterLight's series).
I have tried this, because scanners dont give you manual control and manual override for the basic components of imaging, it's a no-go unless you are a talented electrical engineer.
You may get something like this (pictured: my ceiling light and some hanging film):
Okay so instead of destroying a cheap scanner in the process, we can move on to other methods, the next method is to project an image onto a ground glass or equivalent on the platen, ie: use an enlarging lens (a 35mm film prime in the focal range of 50mm also works as an enlarging lens and can be quite sharp) and project a much larger image over the ground glass (or equivalent) on the platen, for example, we project a 35mm slide onto the ground glass at a size of 140x96mm, now as long as it is focussed correctly and the lens is capable of providing decent resolution we have an effective 4x increase in the scanner's real dpi (and the projected image will be diffraction limited by the effective dpi, say for example your scanner is shit house and is only good for 600 dpi, our effective dpi is then 2400 dpi and the diffraction limit would then be f/19).
This presents some more problems, the backlight needs to be dimmable (or you need to throw some wire mesh/scrim on it to knock the intensity down) in case it floods the scanner with too much light and you end up with a picture of a polar bear in a snow storm.
Secondly, the front light/LEDs in the scanner need to be turned off or they will wash out the image to basically nothing (brings back to the problem of disabling the LEDs and calibration), this isn't a problem if you scanner has a transparency unit, you can scan in a film mode and the light will turn off (the scanner's backlight will come on - but not important).
Thirdly it needs to be light tight from the lens to the platen, so you need to make amake shift light tent around it, not a real problem, just an inconveniance.
And fourthly, you will pick up the grain or texture of the ground glass or equivalent you are projecting onto, now you could try scanning just the ground glass in normal mode and try subtracting in photoshop once you have a sample of it.. but to borrow the technique from a depth of field adapter for digital video cameras.. you can vibrate it with an oscillator, presents a bit of a challenge as it made need to be suspended off the platen glass (or remove the glass and suspend it) to be able to be vibrated suffienctly.. the best solution I can think of would be to suspend it about 1mm off the platen and hope for the best - with little rubber feet used on devices to keep from slipping on surfaces.
Now remember it's just the groung glass that's vibrating/moving, not the projected image, so the grain/texture should essentially get blurred out.
I have yet to try this method, but I will get around to it.
There is one more method which I recently thought of which does not require a ground glass, and it is using a single lens to modify the optical path and where the focus field lies, for example, using a dioptre above the platen which in combination with the scanner's lens modifies the optical system, essentially macro'ing the image, and enlarging the image directly.
I just so happen to have a high quality single lens I dismantled from an old Tamron SP zoom lens.
I will update this as I progress further with this concept.