Repeated scanning generates slightly different images

ady11 · Jul 10, 2024

Hi all,

I have an Epson WF-3540 printer/scanner which I want to use for scanning a large number of mostly 4x6 photo prints.

During some initial tests at 1200 DPI, I noticed that when scanning a picture repeatedly (same position, same scanning parameters), the resulting images are slightly different.

You can see the result here.

Open up one of the images then switch between the two using the left/right arrows. You will see a subtle but visible vertical wave going through the pic, which indicates that the scanning process is not identical between different passes.

Same thing happens at 600 DPI too, but not at 450 DPI or lower.

Is this normal or is there something wrong with my scanner?
Is there something I can do to fix or mitigate this problem?
Is 1200 DPI or even 600 DPI too high for this project? In my tests I noticed that without zooming in, even 300 DPI doesn't look bad, although I would prefer to use a higher res to be sure that the pics are as close to the originals as possible (I intend to get rid of the hard copies once they are scanned).

Thank you.

brbo · Jul 11, 2024

I doubt there is much you can do about that. Just make sure that prints are pushed firmly down on glass bed during scanning.

It could also be that transport mechanism in your scanner that moves the mirror/lens/sensor assembly is not rigid/precise enough. I've seen this happen on other film scanners as well where film transport is not perfectly uniform, like Nortisu lab scanners, Flextights are infamous for this...

koraks · Jul 11, 2024

ady11 said:
Is this normal or is there something wrong with my scanner?

It's probably a limitation of the scanner. What you're seeing is a slight variation in the speed with which the sensor array passes underneath the print. The sensor is moved by a stepper motor and a mechanical transmission, part of which is a drive belt. Imperfections in this system can result in minor variations in step size. The absolute error will likely be comparable in size regardless of step size, which, if you think about it, means that the problem will ultimately be bigger if you scan at higher resolutions, since the ratio between the step size error and actual step size will be bigger. If you find that the problem is gone at 450dpi, then scan at that resolution, or lower. Note that there's very little added value in scanning prints at higher resolutions than 300dpi or so.

ady11 said:
I intend to get rid of the hard copies once they are scanned

That's your call to make, but unless there's a very compelling reason to do so, I'd never recommend to destroy the originals. Too many people have destroyed the negatives and kept the prints, and currently many people are destroying the prints because they have scans. Then a few years down the line they realize what they've done and at that point there's no turning back. Surely, you can afford to keep a box of prints and negatives around. The shelf space it takes up and the times you have to pick it up are a minor inconvenience in the greater scheme of things.

Don_ih · Jul 11, 2024

That Epson all-in-one isn't really the best choice for scanning photos. It was designed for office use (so, for document processing and for speed rather than quality).

koraks said:
Note that there's very little added value in scanning prints at higher resolutions than 300dpi or so.

That depends on the print and intended output. If you're scanning prints that were produced digitally, there's probably no value going to a higher resolution than 400dpi. If you're scanning contact prints or even enlargements, you may want to go significantly higher. The higher the resolution, the better you can apply repairs to damaged photos.

If you scan your 30-year-old photobooth picture at 300dpi, then get it printed 30 inches wide, you'll look like Super Mario unless you're standing a good 100 feet away from the print.

koraks · Jul 11, 2024

Don_ih said:
If you're scanning contact prints or even enlargements, you may want to go significantly higher.

I dunno. I scan prints regularly, especially non-digital ones. I never get much additional data from scan resolutions over 300DPI. Minor effects that go unnoticed to the naked eye start to play a role, such as diffusion due to emulsion thickness, halation in the top layers of the paper, surface unevenness etc. There's just not all that much happening beyond 300dpi on a paper base, contrary to more 'technical' base like PET or glass.

Under ideal circumstances, perhaps there's a slight benefit to scanning at, say, 600dpi instead of 300. You know, wind in the back, going downhill on a good day. The question is, how many of your typical consumer 4x6 prints fall in that category?

Let's take two snapshots from a stack here in my darkroom, one color and one B&W, both RC paper. Both taken with 35mm consumer grade lenses and decently done, but nothing stellar; i.e. no 4x5" Symmar-S contact print or anything. So fairly typical of what OP is probably scanning at this moment. Here's an overview:

This is a 300dpi snippet of both, taken from areas of the image that seem to be advantageous in terms of 'fine' detail:

This is a 1200dpi snippet from the same general areas:

Now compare this to similar snippets taken from the 300dpi scan, but sampled up to fake 1200dpi:

The net gain in real detail is very limited. There is some, I agree, especially in the B&W print, but how meaningful is it to adjust a workflow based on the marginal gains?
I'd much rather preserve the physical originals in order to ascertain that no image data is lost.

The counter-argument obviously would be "but there is some additional data captured" and "digital storage space is cheap". I can see the sense of those arguments.

PS: the scans above are made on an Epson 4990.

bdial · Jul 11, 2024

The variation is normal. On its own, it doesn’t indicate any problem with the scanner.

ady11 · Jul 11, 2024

Thank you all!

bdial said:
The variation is normal. On its own, it doesn’t indicate any problem with the scanner.

Don_ih said:
That Epson all-in-one isn't really the best choice for scanning photos. It was designed for office use (so, for document processing and for speed rather than quality).

I would think that such high res (and it can go even higher than 1200 DPI in pro mode) is mainly for images not documents, but if it can't reproduce them with 100% accuracy (at least from a geometrical if not color standpoint), then I'm not sure what the value of these high res settings is.

Is there a reasonably priced scanner (low $00's) that is better at this than my Epson?

koraks · Jul 11, 2024

ady11 said:
Is there a reasonably priced scanner (low $00's) that is better at this than my Epson?

You could try to find a second-hand 4990 or a later model, like a V600/700 etc. I'd keep an eye on places like facebook marketplace, craigslist etc, see what pops up. The thing that makes scanners expensive is the ability to do transparency/film well, especially large format. But since you're scanning prints, there are likely plenty of more than decent scanners to be had for modest price.

Vaughn · Jul 11, 2024

brbo said:
I doubt there is much you can do about that. Just make sure that prints are pushed firmly down on glass bed during scanning.
...

I agree and suggest that if the machine you are using otherwise is easy enough to use and gives you an adaquate output relative to the source material, keep doing what you are doing! Interesting question!

ady11 · Jul 11, 2024

Vaughn said:
I agree and suggest that if the machine you are using otherwise is easy enough to use and gives you an adaquate output relative to the source material, keep doing what you are doing! Interesting question!

Yes, that's what I would like to do, but the problem is I don't know what "adequate" means for my case. I'm not sure how to choose between resolution and accuracy - ideally I'd like to have both.

brbo · Jul 12, 2024

But, would you ever have noticed the lack of accuracy if you didn't compare two scans of the same print?

Don_ih · Jul 12, 2024

ady11 said:
I don't know what "adequate" means for my case

Compare the scan with the print. If it looks correct, it is probably good enough. And, in this instance, there likely is not much of a reason to go higher than 4-600dpi scanning.

Vaughn · Jul 12, 2024

ady11 said:
Yes, that's what I would like to do, but the problem is I don't know what "adequate" means for my case. I'm not sure how to choose between resolution and accuracy - ideally I'd like to have both.

Why desire that level of "accuraracy?" One has lens distortions, distortions due to film flatness (original capture and scanning) and so on. If the only way one knows there is any distortion is by going back and forth between two scans, the distortion is absolutely insignificant. We are not astrophysicists measuring the distance between stars on a glass plate negative.

reddesert · Jul 13, 2024

I think you can't expect a consumer scanner to have a pixel level of repeatability in the positioning. A small non-repeatability in either the start position of the scan head, or the position as it is driven, will cause small offsets. For ex, at 300 ppi, 1 pixel is 83 microns. It is unlikely that you are going to position a consumer device much more accurately than that in open-loop - mechanisms to position very accurately would typically use encoders to measure the position, much more involved than a consumer grade device. This only matters if you were trying to do accurate measurements off the document/picture, as Vaughn alluded to.

It's my impression that many users feel that the high-dpi specs such as "1200 dpi" on a generic flatbed are advertising numbers. There's not really information at the 20 micron level in a photographic print on paper.

ady11 · Jul 13, 2024

Thank you all. I'm going to experiment with lower res then. That also has the advantage of being much faster, and since I have a lot of prints to scan, it will make a difference.

_T_ · Jul 14, 2024

koraks said:
It's probably a limitation of the scanner. What you're seeing is a slight variation in the speed with which the sensor array passes underneath the print. The sensor is moved by a stepper motor and a mechanical transmission, part of which is a drive belt. Imperfections in this system can result in minor variations in step size. The absolute error will likely be comparable in size regardless of step size, which, if you think about it, means that the problem will ultimately be bigger if you scan at higher resolutions, since the ratio between the step size error and actual step size will be bigger.

I wonder if it would be possible to use the inconsistencies in the movement of the scan head to do something like the pixel shifting that some modern mirrorless cameras do to achieve a higher resolution scan by combining multiple slightly different scans.

ady11 · Jul 14, 2024

That's a good

_T_ said:
I wonder if it would be possible to use the inconsistencies in the movement of the scan head to do something like the pixel shifting that some modern mirrorless cameras do to achieve a higher resolution scan by combining multiple slightly different scans.

That's a good idea, but I think it's beyond my ability unfortunately.

ady11 · Jul 14, 2024

I ran new tests I could see the same distortions even at low resolutions like 300 or 400 DPI (which previously I thought were fine). It looks like I have no choice but to get a new scanner.

Repeated scanning generates slightly different images

ady11

brbo

koraks

Moderator

Don_ih

koraks

Moderator

bdial

ady11

koraks

Moderator

Vaughn

ady11

brbo

Don_ih

Vaughn

reddesert

ady11

_T_

ady11

ady11

Repeated scanning generates slightly different images

Moderator

Moderator

Moderator

Privacy & Transparency

Privacy & Transparency