Is DSLR scanning "better" than flatbed scanning?

lantau · Oct 17, 2019

Researching the topic quite some while ago I found an article (from the 4-8MP era) in a computer magazine. They used a colour correction filter to move the orange masked colour negative into the cameras colour space. I'm doing the same with an 80B Cokin filter below the negative. At least on my system I could see that, after inversion, the blue channel moved more towards the centre. Without the filter it could even clip.

I was recently gifted a Durst M605 enlarger. I put it in storage because I don't want to replace my LPL7700 for the time being. But it has a great glass less negative carrier. I haven't done too much with it yet, but it holds 35mm negatives really flat, and keeps them about 1cm above the surface of the light plate. We discussed in another thread that the camera might pick up the surface structure of the light plate when my negatives where placed on it and covered with AN glass.

That seems to be gone. And with the 80B filter on the plate, 1cm below the negative any scratches in it won't matter any more, either. Of course Adrians method of using flash is the best solution, especially for the reasons he mentioned just above (exclude external light, high shutter speed). But that is too complicated for me at the moment.

I ordered newly made 6x6 masks for the Durst holder (a guy in Italy laser cuts them). Haven't tested them yet, but I really hope it will keep 6x6 negatives fairly flat. Stopping down to f/7.1 with a 30mm macro on a m4/3 camera gives pretty good DoF, though. And since I print only in the Darkroom, a single exposure with 16MP is plenty enough even for medium format.

Helge · Oct 29, 2019

Trouble with not stitching is that (apart from bayer interpolation, problems that take some of what film is truly good at; colour), you run the risk of veiling flare in the lens.
Thats one of the advantages drum scanners still have: A very small aperture that will cut anything but what is being scanned out.
This goes double for 135.

Helge · Oct 29, 2019

Has anyone thought about or even experimented with more advanced grain reduction?
I'm thinking specifically about (the current buzzword for neural networks), deep learning.
Anyone who has seen what it can do for under sampled ray tracing, or image noise in general will immediately have thought about film.

https://blogs.nvidia.com/blog/2017/05/10/ai-for-ray-tracing/
https://www.techspot.com/news/75433-nvidia-ai-can-remove-noise-artifacts-grainy-photos.html

We all know there is a tonne of low contrast, very high resolution information in film, that is somewhat masked by noise, and therefor hard to pull out with naive sharpening algorithms.
Our own neural networks (eye and brain) can pull them out because we "hardwired" for that. That's why film wins if you blow, even a 135 wet print up huge (or just use a projector).
The grain is not always a problem as it sort of can act as a proxy simulacrum of detail, and can act as a stylistic element. It would be nice however to have the option of having access to all the extra detail unmasked.

Are there any commercial plugins or separate applications with above mentioned ideas/technologies yet?

Helge · Oct 29, 2019

Adrian Bacon said:
The other thing to pay attention to is your light source. Here, quality of light, power output, control, and consistency is key. I personally use a strobe. It's very high CRI, I have control over the output power in 1/10 f-stop increments, and it's extremely consistent. More importantly, it has enough power that I can shoot at ISO 100, f/8, and still have a 1/200 shutter speed, effectively eliminating any light contamination from the room and not really having to worry about motion blur or other vibration induced blur.

So you don't use any Wratten filters to "pre-demask" the light for the orange mask in C41?

Adrian Bacon · Oct 29, 2019

Helge said:
Trouble with not stitching is that (apart from bayer interpolation, problems that take some of what film is truly good at; colour), you run the risk of veiling flare in the lens.
Thats one of the advantages drum scanners still have: A very small aperture that will cut anything but what is being scanned out.
This goes double for 135.

That has not been my experience, at least with my setup.

Helge · Oct 29, 2019

Well drumscan expert (? don't know if he would describe himself like that) Tim Parkin said as much.
Having never had access to a drumscanner myself I can't verify. But it's pretty likely when you think about it.
Of course it's not something that is going to be very apparent in all captures.

Adrian Bacon · Oct 29, 2019

Helge said:
So you don't use any Wratten filters to "pre-demask" the light for the orange mask in C41?

Nope. I can, but have never needed to. Even with the orange mask of the film, on the dslr I use, the green channel on the sensor is still the brightest, so I set the exposure so that the film base plus fog for the green channel is just below where the sensor goes non-linear, which on my particular camera is 10,000, so I target the green channel for 8192. On most C-41 emulsions, the red channel usually will end up right below that (again, what camera you use will effect this), and the blue channel will typically be a couple stops down from that in the 2048 range. This gives me ~13 bits of tonal values for the red and green channel, and 10-11 bits of tonal values for the blue channel, which in practice is more than enough. I’ve never seen posterization artifacts as a result of not having enough discrete tone values in the blue channel. This is of course raw, coming off the sensor. Different cameras may have different results, depending on the WB multipliers used to get to neutral for that particular sensor.

Adrian Bacon · Oct 29, 2019

Helge said:
Has anyone thought about or even experimented with more advanced grain reduction?
I'm thinking specifically about (the current buzzword for neural networks), deep learning.
Anyone who has seen what it can do for under sampled ray tracing, or image noise in general will immediately have thought about film.

https://blogs.nvidia.com/blog/2017/05/10/ai-for-ray-tracing/
https://www.techspot.com/news/75433-nvidia-ai-can-remove-noise-artifacts-grainy-photos.html

We all know there is a tonne of low contrast, very high resolution information in film, that is somewhat masked by noise, and therefor hard to pull out with naive sharpening algorithms.
Our own neural networks (eye and brain) can pull them out because we "hardwired" for that. That's why film wins if you blow, even a 135 wet print up huge (or just use a projector).
The grain is not always a problem as it sort of can act as a proxy simulacrum of detail, and can act as a stylistic element. It would be nice however to have the option of having access to all the extra detail unmasked.

Are there any commercial plugins or separate applications with above mentioned ideas/technologies yet?

How much visible grain you have in a scan at reasonable print resolutions is a function of what gamma you’re applying to increase the contrast. In my experience, if you apply just enough to get to linear, grain is typically not objectionable for a normal contrast negative (0.55-0.62 gamma).

What I’ve seen many people do is scan at very high resolutions, then zoom in and go “man, that way too grainy”. ???!!!???!!! Scale it to a nice 11x14 at 300 dpi and print it out, then look at it and see what the grain looks like. For normal negatives, it’s not typically objectionable.

Adrian Bacon · Oct 29, 2019

Helge said:
Well drumscan expert (? don't know if he would describe himself like that) Tim Parkin said as much.
Having never had access to a drumscanner myself I can't verify. But it's pretty likely when you think about it.
Of course it's not something that is going to be very apparent in all captures.

I’d only take what a drum scan expert would say about dslr scanning seriously if he was also a dslr scanning expert, otherwise it’s just conjecture, and oftentimes just somebody on the internet asserting something to support their chosen method. I wouldn’t call myself a dslr scanning expert, however I’ve probably scanned more film that way than most other people around these parts of the internet. I very rarely say anything about drum scanning because I’m not an expert in that area. I do know that dslr scanning can provide more than adequate results for the vast majority of uses, and is very reliable and fast.

Helge · Oct 29, 2019

Adrian Bacon said:
I’d only take what a drum scan expert would say about dslr scanning seriously if he was also a dslr scanning expert, otherwise it’s just conjecture, and oftentimes just somebody on the internet asserting something to support their chosen method. I wouldn’t call myself a dslr scanning expert, however I’ve probably scanned more film that way than most other people around these parts of the internet. I very rarely say anything about drum scanning because I’m not an expert in that area. I do know that dslr scanning can provide more than adequate results for the vast majority of uses, and is very reliable and fast.

Well, he actually advocated for DSLR scanning for 135 film
Tim Parkin is a good guy, and while he probably would not describe himself as "expert" he knows more than 99.9 percent of people waxing lyrical on the subject of scanning.
Do a search if you don't know him. He is on here too.

Adrian Bacon · Oct 29, 2019

Helge said:
Well, he actually advocated for DSLR scanning for 135 film
Tim Parkin is a good guy, and while he probably would not describe himself as "expert" he knows more than 99.9 percent of people waxing lyrical on the subject of scanning.
Do a search if you don't know him. He is on here too.

Can’t say I know him, and my intent isn’t to discredit. Personally, I’m happy to answer questions other people ask and dispense information based on my own experience of scanning very large amounts of film for other people using my chosen method, which has been honed and refined through lots and lots of scanning.

Helge · Oct 29, 2019

Adrian Bacon said:
How much visible grain you have in a scan at reasonable print resolutions is a function of what gamma you’re applying to increase the contrast. In my experience, if you apply just enough to get to linear, grain is typically not objectionable for a normal contrast negative (0.55-0.62 gamma).

What I’ve seen many people do is scan at very high resolutions, then zoom in and go “man, that way too grainy”. ???!!!???!!! Scale it to a nice 11x14 at 300 dpi and print it out, then look at it and see what the grain looks like. For normal negatives, it’s not typically objectionable.

Well that's just plain simple filtering.
My question pertained to how to get all the potential resolution of film, which is significantly more than most people imagine from their experience with subpar scanners.
And in particular towards the exciting new development in neural net filtering of images.
As said, we as humans can easily "see through" the grain, but it's still there, and is sometimes unwanted.

Alan Edward Klein · Oct 29, 2019

Does anyone know how Epsonscan handles the orange mask in their scanners?

Alan Edward Klein · Oct 29, 2019

COuld you apply their methods to camera scans?

Helge · Oct 29, 2019

Alan Edward Klein said:
COuld you apply their methods to camera scans?

Perhaps you could trick the software to import images files other than from the scanner?

Adrian Bacon · Oct 29, 2019

Helge said:
Well that's just plain simple filtering.
My question pertained to how to get all the potential resolution of film, which is significantly more than most people imagine from their experience with subpar scanners.
And in particular towards the exciting new development in neural net filtering of images.
As said, we as humans can easily "see through" the grain, but it's still there, and is sometimes unwanted.

All the potential resolution of any image is a combination of the film used along with the shutter speed, lens, focus, and aperture used to take the picture, and is often times much lower than what most people realize. To look at film resolution alone is folly. Sure, given the technology available, you can scan down to the particle level, however the point where usable image resolution stops is more a function of everything else. Don’t get me wrong, I’m generally all for more resolution when scanning, however, in my experience, there absolutely is a point of diminishing returns where adding more resolution does not appreciably make a given image in a given format look better in terms of resolution.

I personally prefer to live in the “good enough for the vast majority of uses” realm as a matter of productive practicality.

Helge · Oct 29, 2019

Absolutely, but there is still that rare case when you want to print big.
Looking at projected slide, on a quality projector you can easily see the huge potential in film for large prints, without ever printing that big.

Also film runs into problems with grain aliasing at too low resolutions whether the complete image chain warrants the actual resolution or not.
This contributes to the bad rep film gets sometimes as excessively grainy and noisy.

If you use a tripod, have optimal light, optimal aperture and use an even half decent prime lens (FD 50mm 1.4, Zeiss 50mm 1.7 or Nikkor AIs 28mm 2.8 for example) , it's not that hard to get over the often sited 24MP max resolution of 135 film.

Adrian Bacon · Oct 29, 2019

Helge said:
Absolutely, but there is still that rare case when you want to print big.
Looking at projected slide, on a quality projector you can easily see the huge potential in film for large prints, without ever printing that big.

Also film runs into problems with grain aliasing at too low resolutions if the complete image chain warrants the actual resolution or not. This contributes to the bad rep film gets sometimes as excessively grainy and noisy.

If you use a tripod, have optimal light, optimal aperture and use an even half decent good prime lens (FD 50mm 1.4, Zeiss 50mm 1.7 or Nikkor AIs 28mm 2.8 for example) , it's not that hard to get over the often sited 24MP max resolution of film.

No argument from me there, though I’d say in those instances where you want/need more resolution, you’ll get a lot more bang for the buck by going to a larger film format.

Helge · Oct 29, 2019

You might want the small lenses that accompany the smaller format.
An equivalent to even a 200mm f4 is huge in 6x6. Same with wide lenses. They're big and worse yet slow, compared to what similar money (if even available) will buy you on 135.
A medium format folder is the only instance where you actually get a very portable camera, with larger film.

GLS · Oct 30, 2019

Helge said:
Same with wide lenses. They're big and worse yet slow, compared to what similar money (if even available) will buy you on 135.

A counter argument (other than the obvious resolution differences) is that the perspective distortion you get with 135 format wide angles is far more noticeable and, IMO, very distracting most of the time. Since getting used to shooting 6x6 and 6x7 I have started to really dislike the 135 "wide look".

Besides, an f4-ish lens isn't that slow.

Adrian Bacon · Oct 30, 2019

GLS said:
A counter argument (other than the obvious resolution differences) is that the perspective distortion you get with 135 format wide angles is far more noticeable and, IMO, very distracting most of the time. Since getting used to shooting 6x6 and 6x7 I have started to really dislike the 135 "wide look".

Besides, an f4-ish lens isn't that slow.

And you can shoot 800 speed film in the larger 120 roll format with a lot less grain penalty. My Fuji GW690 isn’t that big and it spits out negatives that are absolutely gorgeous, and have the same aspect ratio as 135. It is literally like a quad resolution 35mm camera.

John51 · Oct 30, 2019

For dslr scanning, I'm thinking on the lines of those slide copiers from the 70s. Attach it to the camera and it's ready to work. A 24mp sensor could do a 35mm neg in one shot. A neg carrier with indexing could do a 120 neg in say 6 shots.

Helge · Oct 30, 2019

John51 said:
For dslr scanning, I'm thinking on the lines of those slide copiers from the 70s. Attach it to the camera and it's ready to work. A 24mp sensor could do a 35mm neg in one shot. A neg carrier with indexing could do a 120 neg in say 6 shots.

As said folders are the exception. But you are “stuck” with that lens and often rangefinder that could be out of calibration.
And 6x9 is over five times the potential resolution of 24x36. Almost half a 4x5.
No one is contesting that it’s an awesome format. But then again it’s a question of, how often do you really need the resolution?

Helge · Oct 30, 2019

GLS said:
A counter argument (other than the obvious resolution differences) is that the perspective distortion you get with 135 format wide angles is far more noticeable and, IMO, very distracting most of the time. Since getting used to shooting 6x6 and 6x7 I have started to really dislike the 135 "wide look".

Besides, an f4-ish lens isn't that slow.

Absolutely. Medium format is wonderful, and I shoot it all the time.
I just wouldn’t want to lug a medium format SLR, a set of primes and a tripod around all day.
For location shooting, where you know exactly where and what, it’s no contest medium format is better in more ways than just resolution.

GLS · Oct 30, 2019

John51 said:
A neg carrier with indexing could do a 120 neg in say 6 shots

Does such a contraption exist as a commercial product? I would be very interested if so.

Is DSLR scanning "better" than flatbed scanning?

Privacy & Transparency

Privacy & Transparency