If you are sitting on a horde of negatives, waiting for the digital photography fad to die off, it may be time to think about digitizing your old film. [Kinpro1024] can help with the PiDigitzier, an open-source film scanning solution. The build centers around a Pi Zero 2, a Pi HQ camera, and a diffusing LED lighting fixture. Of course, there’s also some miscellaneous hardware and a camera lens; the example used a Pentax 50 mm f1.8 lens.
Half of the project is mechanical. An MDF tower provides a stable 250 mm workspace and decks that can slide up and down using threaded rods and curtain rods. Apparently, leveling the platforms is important not only for the optics but also to allow the MDF to move along the rods without binding.
On the software side, some Python software orchestrates the capture and builds a high-resolution scan by combining three different images from different positions. As you might expect, though, the image stitching doesn’t occur on the Pi.
[Kinpro1024] has only scanned black and white film so far because the LEDs have a poor color rendering index. If you wanted to scan color film, you may have to experiment with better lighting or software correction. The camera’s rolling shutter also causes some issues if the LEDs don’t hold their intensity through the entire scan. You could, of course, replace this with a global shutter camera.
Stray light can also be a problem. Ideally, you’d use bellows like a big camera, but if that’s not available, you can wing it with some homemade light baffles.
If you decide you want to try color, you will be interested in this project. You might also think of raiding the kids’ Lego box.
I’ve never understood the concern with high CRI for scanning film. Color film is essentially an RGB technology. Ideally you’d use an RGB backlight with wavelengths chosen that give you the strongest contrast between color channels – ie pick a color of red that your red (cyan) layer blocks most strongly and your green (magenta) and blue (yellow) layers pass through most unaffected. Pick a green and blue wavelength similarly, combine together into an RGB backlight, and you’ve got an ideal backlight for film scanning which would have an awful CRI.
Chances are, any random narrow-band RGB LED is closer to that ideal than a wideband white light.
There are films with extra color layers, like Fuji Superia 200, that might work better with 4-wavelength illumination or broadband illumination.
And what happens when your film/sensor combination doesn’t work with that narrow band light very well this time? High CRI = reliably work with anything, and usually the scan comes off the machine looking good with basically no post processing, where your system you are going to have put in effort to get every colour channel to be roughly right for every single roll of film no doubt…
In theory high CRI lighting is about to get a lot cheaper, we’ll see how long things take to come to market.
In the meantime building your own light sources isn’t actually a bad idea, even for this purpose, s long as it is tunable. Film spectral sensitivity isn’t linear, the information it carries shifts with age to a degree, (also non-linearly) and every vendor is different.
e.g. a simple high CRI scan will give you a “good” picture, but not the best. It’s our job to throw together something fun that can, ideally with some degree of automation.
Could you please elaborate on high CRI getting cheaper?
Tungsten, or whatever other black body radiator light source, can be pretty cheap, but back in the 1970s narrowband RGB lighting for color film reproduction was worth spending a lot of money on. Minolta had a gas discharge tube enlarging head that was coveted.
Once you have something on film, having more that three discrete wavelengths of light just makes things more difficult. Even lens design is easier without correcting for the in between wavelengths.
When film dyes fade, the wavelengths don’t shift, the density fades.
Thanks, Hugh, I appreciate the answer but I was more specifically asking what S O meant by
“In theory high CRI lighting is about to get a lot cheaper, we’ll see how long things take to come to market.”
The problem is, you’re mixing things up: Back then, the RGB lights were used for photo-chemical reproduction! The chemistry was adjusted for this process and vice versa. Therefore of you want to reproduce a negative (e.g. as a positive print onto Photo-sensitive paper), RGB IS STILL A THING. But: even though modern displays still work with RGB light sources, the sensors have not the same sensitivity to the different wavelengths as the photo chemistry. They’re basically monochrome sensors, which get color filters on them and artificially create color bases on which filter lets through how much light. This technology still works best with natural light of a broad spectrum, because their whole internal colour science (which is software, not chemistry) is designed for this. You always have to adjust your light source for the medium onto which you want to reproduce it to.
Easy example: You always shot Portraits on film with studio flash lights. Your system worked well. Would it still make sense to use those studio flash lights if your sensor is made for Infrared? No, you use IR lightsources instead.
I don’t know much about photography. But I had a lot of trouble with LED backlights for LCDs we chose. We were trying to get a standard color gamut (rec.709) out of our displays, but it just wasn’t going to happen with the parts in the BOM. I would have love a 95 CRI white light source instead of having to deal with manufacturing variations on the RGB modules we bought. But price ultimately decided the fate of that device and we implemented some soft clipping so it doesn’t look so obviously bad.
Any reduction of the Spectrum will lead to poor results. If you had a green surface with a narrow band reflection spectrum at 550 nm and you’d shine a green 532 nm laser on it, your green colored surface will appear black, even if both the surface and light source appear almost the same color! That is the explanation.
cri is not the problem, but lack of contrast is. a bw negative has a decent contrast, but colour negatives don’t. the camera module is configured for a 6 stop contrast difference, but a colour negative only has about 2 or 3 stops. so if you take a 8 bit camera, only about 5 bits are used. the resultic “scan” will miss a lot of colour and contrast info.
but then again, something is better than nothing.
Every film scanner I’ve used has a second pass with an IR LED to capture a map of surface dust to remove from the final image.
This is way less common than you imply. Most “film scanners” are just a tray on a conventional scanner, sometimes even an inferior one.
Scanning anything is more nuanced than this. Those stops are not a hard limit. Among other things you can tease out A bit more information related to the actual spectral sensitivity of the film, which isn’t linear either. It’s not a digital medium, don’t treat it like one.
Gaah! My internet connection just ate a carefully researched and written reply.
I’ll summarize. Kodak color negative films have a density range of about log10(100) = 2, or 6.65 bits.
Some cameras encode JPEG nonlinearly, which may help ameliorate the intensity gap from an 8 bit camera.
Using a camera of more than 8 bits would help. A scanner with a light source blue stronger than green stronger than red would help, compensating for the orange color of negatives.
I scanned 50 years worth of my parents’ color slides. The ones on Fujifilm were fine but on some of the cheap no-brand slides the colors were wrong – often red had faded or was missing. I discovered that converting those slides to b&w on the computer made them look perfectly presentable again {but b&w of course.)
Mentioning this in case anyone else encounters this – if I had not discovered the b&w trick quite a few would have been lost.
If only certain spectra are weak, you can try selectively enhancing the contrast of just that colour. Note I didn’t say “channel” because we’re talking about film chemistry. It’s likely the loss of red isn’t the only thing wrong. For a quick hack just try out manipulating the red colour channel to see if there is anything there in your scans.
We scanned my parent’s slides, dating back to the early ’50s. The Kodachrome ones were fine, but The Ektachrome (and similar processes from other companies) were badly faded. It’s interesting that your Fuji ones were still fine.
That’s interesting. As a contrary note, Kodachrome faded more rapidly in proportion to the amount of light passing through it.
I’ve never liked “one-shot” film scanners that use a single “diffused” light source, because that light is never diffused enough. There’s always a “hot spot” brighter in the center than the edges. If a pure gray negative–or no negative–is scanned and brightness plotted as a 3D height, you’ll see a “hill” at the center. At any rate it’s not flat. Theoretically you could do a “flat-field” compensation to the scan like NASA does, but I’ve never been able to get it to work. I prefer a flatbed-type scanner: an evenly-lit 1-D sensor light that scans down the negative building the 2D image at the cost of time.
At least in the SF Bay Area, the local libraries have recently gotten flatbed slide and negative scanners, so good film scanning is now available free to the public as long as as you get in line.
Very few film or photo scanners do that, even high end drum scanners for film use a thin emitter and sensor to avoid this very problem.
The lens itself has a “hot spot” no matter how even the light source is. That’s simply how lenses work. Chemical photography equipment and photographers compensate for that.
I was very happy with my “diffused” light source and compared to the quality of the old negatives or sildes it did not really matter ;-)
In my case the results were very good. (https://hackaday.io/project/174961-one-week-build-slides-digitizer)
The lenses in cameras will have falloff toward the edges. If you’re scanning negatives with a scanner that has a hot spot problem, there will be some compensation of the camera’s problem by the scanner’s problem.
You’re all over-thinking this. The digital camera sensor itself has RGB microlens filters with a fairly narrow bandwidth. They’re fixed, and additional RGB filtering is unnecessary and might even degrade the results.
I’ve scanned hundreds of slides and colour negatives, using a variety of methods and light sources – flash, white LEDs, etc. None of them resulted in colour that couldn’t be well corrected with a white-balance tweak.
Colour negatives are a category all to themselves though. Through trial and error I’ve found that negating the orange mask with a physical (optical) blue filter makes the reversal and channel balance much, much easier and straightforward.
You can ‘get rid’ of the mask using the camera white-balance, but you lose digital bit-depth in the capture by doing that.
Anyhow, the subject of camera ‘scanning’ has been discussed and beaten to death on many photography fora: Notably in Photo.net where it has its own thread under Scanning and Scanners –
https://www.photo.net/forums/forum/55-scanning-scanners/