"Scanning" negatives with digital camera on a light table

[J N]

Honestly, I don't think there is any point in using my V850 for anything other than reflective scanning any longer, and I hate myself for not changing my workflow many years ago. I put my negatives on a well diffused high CRI light table and photograph them with my a7iii.

Super dense negative? No problem. Expose longer.

Super dense and contrasty? No problem. Resort to bracketing and HDR.

Super fogged? No problem. Expose longer.

It's also faster. Can be much faster depending what kind of jig you have set up. You can photograph your sprocket holes and edge markings if you like.

The only issue is that you'll be working with some reversed controls, at least in LR, even if you have a separate homemade "negative" profile. But that's a minor thing. Even focus is less of an issue as you can crank your macro lens up to f/11 or f/16, and you can preview focus as you shoot. Resolution isn't an issue unless you are working with large format.

Just my $0.02.

 

[Romanko]

Full-frame digital camera, a good macro lens, light-table, copy-stand and film holders cost much more than a flat-bed scanner. Otherwise, camera scanning seems a better solution.

Resolution isn't an issue unless you are working with large format

With large format negatives you can get ridiculous resolution (gigapixels) if you use stitching and pixel shift.

 

[Alan Edward Klein]

I'd have to buy a digital camera beside my P&S and micro 4/3. How much for the whole setup? What do you use?

 

[250swb]

Think of it like you are buying a very high end scanner but you also get a full frame (or pixel shift) camera thrown in for free! All the other things, like a light pad, macro lens, copy stand, and neg holder are all enjoyably nerdy to accumulate and talk about, and that is the spirit this forum is built on.

 

[JerseyDoug]

If you already have a digital camera and a macro lens that will focus to 1:1 the VALOI easy35 for €229 is a complete solution for scanning 35mm, replacing the copy stand, light source and negative holder.

It's what I am using now with a Fuji X-T20 (<$500 on eBay) and a 40/4 EL NIKKOR enlarging lens (<$50 on eBay).

 

[Romanko]

You can get away with a relatively inexpensive setup. Kyle McDougall made a good video on this:

 

[GLS]

With large format negatives you can get ridiculous resolution (gigapixels) if you use stitching and pixel shift.

True but working with such files becomes impractical to put it mildly, especially with layers.

 

[Romanko]

True but working with such files becomes impractical to put it mildly, especially with layers.

I doubt that technical problems could stop people who shoot large-format cameras. This would certainly require a different workflow. The images could be processed individually before stitching them together.

 

[_T_]

You can’t do very much processing on the pre stitched images or the colors won’t match once they’re stitched.

The limitations you’ll hit with such large files are the ram and swap disk space. With such large files most computers are going to run out of ram immediately, but the more you have the faster you can work on your images. The size of file you can handle will usually be limited by the capacity of your swap disk and you’ll see speed improvements with faster storage.

But a 4x5 negative holds somewhere around the equivalent of a minimum of 250Mp of data depending on the film. With my epson v600 I’m able to get something like 130Mp out of each sheet and I can’t imagine needing more than that. At 130Mp there’s already no limitation on what size prints you can make. It’s such overkill for viewing on any screen of any size that it’s basically pointless to make a file that big if you’re not going to print them.

With 130Mp it’s manageable to work with 16GB of memory and an nvme swap drive with sufficient space.

 

[wiltw]

Converting the digital image of a color neg to positive image also takes SOFTWARE that is designed to do that job with reasonable quality.
I tried to take a neg and convert it myself without special conversion software as an experiment, and this is what resulted...

The negative photographed with camera

The negative image inverted with ordinary postprocessing software

The inverted image with attempted changes to alter color balance

The same negative scanned on a flatbed scanner and converted with the scanner software....looks much better, and remarkably similar to the original color print!

 

[Romanko]

You can’t do very much processing on the pre stitched images or the colors won’t match once they’re stitched.

Software capable of batch processing can solve this problem.

The limitations you’ll hit with such large files are the ram and swap disk space.

I don't know how the photo-editing software works but I would assume it loads the whole image into memory. Technically, it is possible to read and write image files in chunks thus reducing the memory footprint to whatever RAM is available.

The question of why anyone would go into all this trouble still remains. Suppose you want a giant 8 x 10 meter print. You won't be viewing it from a 30 cm distance thus you don't need 300 dpi resolution. You can do the maths and estimate the required resolution, it will be in the order of tens of dpi if not less.


So, can we do super-resolution scans? Probably. Do we need them? Probably not.

 

[Romanko]

I tried to take a neg and convert it myself without special conversion software as an experiment, and this is what resulted...

I feel your pain. I've been trying for several years now and still nowhere close. Apparently, very good results could be achieved by using just the curves function available even in the most basic software. Have a look at this thread:

Exposure/Filters/Flash Exposure Assistance (Medium Format Film)

Hello community! I hope this is in the right place to help an amateur hobbyist in medium format film. I went to Japan with my medium format camera to enjoy time with my family. I have been reviewing my photos and exposures. I believe I know what went wrong, but would appreciate further...

www.photrio.com

It won't be perfect but you should be able to get much closer to the image produced with scanning software.

 

[koraks]

Technically, it is possible to read and write image files in chunks thus reducing the memory footprint to whatever RAM is available.

That's basically what virtual memory in combination with a pagefile ('swapping') does. It can't just read chunks from a file one at a time because most file types aren't literal representations of the color channels (BMP and certain flavors of uncompressed TIFF are exceptions), so instead, the editing software loads the entire file, decodes it into a literal pixel map and commits this to virtual memory. The operating system then handles the relationship between hardware RAM and the pagefile, although higher end software like Photoshop and GIMP will take over part of this job for optimization purposes. Either way, what you propose is fairly close to how it was envisioned back in the...1960s I suppose? Albeit that even back then, ideas about decoupling between different layers (hardware, OS, application) were already more advanced.

So, can we do super-resolution scans? Probably. Do we need them? Probably not.

Yeah. I agree, and also with @_T_ 's 130Mpix limit. I personally have never had the need for anything beyond 100MPix and even that was stretching it by a huge margin. I'd love to be one of the cool guys who routinely prints murals several feet across that are displayed in museums for up-close-and-personal perusal of art aficionados. But in reality, I'm not, and only very few of my images ever make it beyond 8x10. I did some approx. 16x20 inkjet prints from 6x6cm and 8x10" negatives the other day, flatbed scanned, and the resolution was ample.

There comes a point where it really does make sense to ask questions about realistic requirements on the entire imaging chain. And I guess for the once in a blue moon very big print, a custom solution can be thought of, such as farming out that one scan to someone like @dokko, and get a resolution that blows anything out of the water that mere mortals can purchase.

 

[_T_]

Software capable of batch processing can solve this problem.

You can make some changes to your image that way, but it doesn’t really extend the capabilities of what you can accomplish when working on only a portion of the image at a time.

The problem is that what you think looks good for one portion might look wrong for another. It’s very difficult to know what’s going to work without being able to see how it affects to the entire image at once.

You could get around this for some things by working on a low resolution version of the file, but once you start using any kind of masks you’re not really able to avoid the heavy lifting because you’re either going to have to upscale the masks or create new masks for the larger files and that’s going to be the most intensive part you encounter besides the stitching itself.

All the other stuff you might typically do is not that bad. Just don’t try to run any of the cpu intensive filters because they will just crash.

 

[GLS]

Once cropped to the frame edges I end up with images from 4x5 of around 190 MP from the A7RIV's pixel-shift capture. At 240 ppi this would produce a native print size 64 inches across. I cannot imagine ever printing this large.

 

[MattKing]

If, before inversion, you create a layer that is just the mask plus clear film, and then divide out that layer, and then collapse the layers and then invert, it works better.
The terminology varies with the software you use, but the approach is essentially to first neutralize the colour of the negative before you do the inversion.

 

[brbo]

If you use pixel-shift and end up with huge files, it pays to have a way to establish how much more true resolution you actually gain by pixel-shifting.

On the m4/3 (Olympus E-M5II) I find that pixel-shift is less effective than theory would suggest (doubling of linear resolution) no matter how good a lens I use. By using pixel-shift I can get measure about 4.000dpi (vs. 2.800dpi without pixel-shift) real resolution with USAF '51 resolution target. So my 60MP pixel-shifted files only carry about 20MP of true data. Resize 60MP to 20MP and you are saving huge amounts of memory (disk and ram) with no real loss in image quality.

 

[titrisol]

It is possible, and it works fine.
Just use the camera you have and adjust the mag ratio for your camera
I use a Pentax camera (APSC 0.6x for 35mm and 0.35x for 6x6) with an old Bellows+Neg duplicator for 35mm and a 50mm macro and enlarger neg holders for 645 and 66
It takes some practice and for me the challenge was getting the bellows close enough for the small mag ratios
For software I started using darktable+negadoctor for inverting and have had good results but as usual YMMV

 

[wiltw]

I feel your pain. I've been trying for several years now and still nowhere close. Apparently, very good results could be achieved by using just the curves function available even in the most basic software. Have a look at this thread:

Exposure/Filters/Flash Exposure Assistance (Medium Format Film)

Hello community! I hope this is in the right place to help an amateur hobbyist in medium format film. I went to Japan with my medium format camera to enjoy time with my family. I have been reviewing my photos and exposures. I believe I know what went wrong, but would appreciate further...

www.photrio.com

It won't be perfect but you should be able to get much closer to the image produced with scanning software.

Someone has taken my JPG of the negative and utilized software (IIRC, 'Negative Lab Pro') to convert the image, and its result was pretty good...better than my outcome using the capabilities within PaintShop Pro to do the conversion...but I think not as good a result as my scanner software conversion!

 

[GLS]

Someone has taken my JPG of the negative and utilized software (IIRC, 'Negative Lab Pro') to convert the image, and its result was pretty good...better than my outcome using the capabilities within PaintShop Pro to do the convertion...but I think not as good a result as my scanner software conversion!

Inverting a jpg copy is never going to give an optimal result, so I wouldn't judge the software based on that. Ideally a raw format should be used, or a 16-bit TIFF. The nature of the light source used to digitise is also important.

 

[wiltw]

Inverting a jpg copy is never going to give an optimal result, so I wouldn't judge the software based on that. Ideally a raw format should be used, or a 16-bit TIFF. The nature of the light source used to digitise is also important.

I shot the neg using real daylight backlighting. I guess I should also have shot with 3400K to more closely mimic the source in a color enlarger head.
I did not attempt RAW file, as I considered the RAW conversion of Paintshop Pro to be inferior compared to Lightroom (conclusion arrived at via comparative tests done years ago). I need to investigate to see what capability for inversion exists within Lightroom. As I indicated earlier, it was a quick and dirty test to see how simplistic the process of inversion might be (or not).
As Romanko states, " I've been trying for several years now and still nowhere close." indicates further experimentation can prove non productive.


Matt King made a statement about 'masking before inversion'...did Romanko attempt that process? And how labor intenstive is it to create said mask and use the process to 'divide out' that layer (estimated time per image added to complete these steps)? While that might well work for a single image, lots of folks have hundreds (or more) negs on file to convert to digital form, so the time investment is a consideration. At some point, it might be better to buy the special conversion software!

 

[koraks]

I guess I should also have shot with 3400K to more closely mimic the source in a color enlarger head.

I don't think that's the rationale, no. But there are a couple of philosophies on illumination for camera scanning. I guess the dichroic head philosophy is one. There's a fairly recent thread here somewhere where we go pretty deep into it. I've also written a blog on it.

The problem with your jpeg is that you smash everything into 8 bit upon capture and then after inversion you start to stretch out some tiny peaks of the histogram where the color information happens to be. This results in massive posterization. If you start with something closer to 12-16 bit, you have orders of magnitude more color information to work with, so that you may ultimately end up with something presentable in 8-10 bit space.

 

[Romanko]

As Romanko states, " I've been trying for several years now and still nowhere close." indicates further experimentation can prove non productive.

My requirements are very specific. I sometimes scan old and degraded color negative films. I want to be able to both extract as much information from the negative as possible (information intent) and preserve the look of degraded film (perceptual intent). Negative conversion software like Darktable Negadoctor and Negative Lab Pro is good for the former task. With a bit of practice you can produce acceptable results fairly soon, so do don't be discouraged by my comments. It is the second task that requires special scanning techniques and software.

Matt King made a statement about 'masking before inversion'...did Romanko attempt that process?

No. I tried pixel shift only once on a 35 mm slide film. The scan did not require a lot of post-processing. There was no practical reason for getting a 200 MP scan, I did it out of curiousity.

 

[Romanko]

But there are a couple of philosophies on illumination for camera scanning. I guess the dichroic head philosophy is one. There's a fairly recent thread here somewhere where we go pretty deep into it. I've also written a blog on it.

Could you please share the links? I often struggle with finding things in this huge forum.

 

[koraks]

Could you please share the links? I often struggle with finding things in this huge forum.

Can't blame you for it, either. See here: https://www.photrio.com/forum/threads/scanning-film-with-digital-camera.199916/ I'd skip ahead to page 2 where we start discussing light sources, and things start in earnest around post #36. There's an excellent comparison in #44 that you may want to have a look at.

The matter was discussed a couple times more since then; I think I came across a link to another blog post on this recently, but I don't recall the details presently.