Scanner comparison: Epson V750 Pro added

[Svitantti]

I think the color balance and contrast could be just visually matched to achieve somewhat comparable results for the noise levels etc. Of course to do a really scientific test, some other method would be better.

I've never noticed problems coming from curved film, but I dont print from scans, I do that it in the darkroom. I scan only for web photos, but I observe the scans quite critically to see what can I get in the darkroom and I think a good resolution helps making good-looking web photos too, even though much of the resolution is wasted.

I think the results are easily acceptable even with the Epson holders without special care of film flatness. Even V700 (which I guess is one of the best regular film scanning flatbeds) is soft whatever you do. Film flatness would be more critical when there actually was sharp areas and others that were blurred (clearly misfocused).

I am no expert on drum scanners, but I think the advantages appear much also when comparing noise to more basic scanners. Less noise means better shadow detail. The (real) resolution also is better, but this is probably because the drum scanners are expensive and well made, with good optics etc. (and not just because they are drum scanners).

[sanking]

Compared to any consumer flatbed drum scanners are incredibly well made and precise, as are the high end flatbeds like Cezanne, Fuji and EverSmart. If you have ever worked with either a drum or high end flatbed you immediately appreciate the precision.

BTW, drum scanners do not use lenses. Rather, they make a direct sample of the area being scanner (or perhaps thousands of samples). This bypasses problems introduced by lenses.

However, the high end flatbeds are very good scanners. I have had the opportunity to compare scans done on a number of drums scanners and my EverSmart Pro flatbed, and the EverSmart compares very favorably to drum scans done at 4000spi to 5000 spi, even though its optical resolution is only 3175 spi. One would have to make a huge print to see any difference.


Sandy King

I am no expert on drum scanners, but I think the advantages appear much also when comparing noise to more basic scanners. Less noise means better shadow detail. The (real) resolution also is better, but this is probably because the drum scanners are expensive and well made, with good optics etc. (and not just because they are drum scanners).

[harrykauf]

Just a quick explanation of the noise removal for the V750 scans.
Since the noise has mainly green and magenta components I use the
a channel in Lab on top of the L channel in hard light mode.
Here you can see the reason why it works:

http://www.sooshee.com/tmp/abney/noise_big.jpg

the a channel stores the green/magenta values of an image (green in dark,
magenta in bright, neutral is in the middle)
Since the noise is per pixel you get an isolated version of the noise without
the luma component, the actual detail of the image.
So you use the reversed version of the chroma noise to fill the "gaps" in the
Lightness channel without affecting the image detail much.

The easiest way to do that is just to copy the L channel and the a into a
new grayscale file, combine them in hard light mode and copy the result
back into the L channel.

[Keefe Borden]

This has been an excellent discussion about scanning even if I don't understand everything here and even if not everyone agrees on the results as presented. There are a couple of questions that (as far as I can tell) have never been addressed. First, is there a point at which you get diminishing returns from a given film? In other words, is there a point at which film simply cannot provide additional detail regardless of the quality of the scanner? If so, does anyone have an opinion about where that point starts?

Here's a related question that drives at the same point: if I start with a 6 X 7 negative and scan at 4,620 spi, I will get a digital file of approximately 10,000 X 12,550 pixels. If I take a 4 X 5 negative and scan at 2,515 spi, I will get a digital file of approximately the same dimensions. (These are approximations, by the way). Assuming the lenses used are of equal quality and all else is held equal, will the print from the 6 X 7 yield the same detail as a print from the 4 X 5 negative?

I can accept that film scanned at 20 spi probably gives twice the detail scanned at 10spi and that 40 spi probably provides twice the detail of film scanned at 20 spi. But does 4,000 spi give twice the detail of 2,000 spi? Does 6,000 spi yield twice the detail of 3,000 spi?

Thanks in advance for all insight.
Best, Keefe.

[sanking]

Good questions.

With regard to the first question, “is there a point at which you get diminishing returns from a given film? In other words, is there a point at which film simply cannot provide additional detail regardless of the quality of the scanner?”

My opinion on this is that the camera/lens systes normally provides the limit to detail. Very simple concept in my opinion, -- if the detail is not on the negative it does not matter what kind of scanner you use, you can not pull detail out if there is none there. If in doubt, look at the detail on the negative with a microscope at 30X-40X.

Work backward to determine the limits of the scan. If you determine that your camera system can only provide a maximum of 60 l/mm (which is extremely high for 4X5), then there would never be a need to scan at more than 3000 spi. In fact, most 4X5 negatives don’t have even half that amount of equivalent detail.

Basically, if you scan at a resolution beyond the limit of the detail in the film you get nothing in return but larger file size. But you can brag about the really huge file size.





Sandy King


Here's a related question that drives at the same point: if I start with a 6 X 7 negative and scan at 4,620 spi, I will get a digital file of approximately 10,000 X 12,550 pixels. If I take a 4 X 5 negative and scan at 2,515 spi, I will get a digital file of approximately the same dimensions. (These are approximations, by the way). Assuming the lenses used are of equal quality and all else is held equal, will the print from the 6 X 7 yield the same detail as a print from the 4 X 5 negative?

I can accept that film scanned at 20 spi probably gives twice the detail scanned at 10spi and that 40 spi probably provides twice the detail of film scanned at 20 spi. But does 4,000 spi give twice the detail of 2,000 spi? Does 6,000 spi yield twice the detail of 3,000 spi?

Thanks in advance for all insight.
Best, Keefe.

This has been an excellent discussion about scanning even if I don't understand everything here and even if not everyone agrees on the results as presented. There are a couple of questions that (as far as I can tell) have never been addressed. First, is there a point at which you get diminishing returns from a given film? In other words, is there a point at which film simply cannot provide additional detail regardless of the quality of the scanner? If so, does anyone have an opinion about where that point starts?

Here's a related question that drives at the same point: if I start with a 6 X 7 negative and scan at 4,620 spi, I will get a digital file of approximately 10,000 X 12,550 pixels. If I take a 4 X 5 negative and scan at 2,515 spi, I will get a digital file of approximately the same dimensions. (These are approximations, by the way). Assuming the lenses used are of equal quality and all else is held equal, will the print from the 6 X 7 yield the same detail as a print from the 4 X 5 negative?

I can accept that film scanned at 20 spi probably gives twice the detail scanned at 10spi and that 40 spi probably provides twice the detail of film scanned at 20 spi. But does 4,000 spi give twice the detail of 2,000 spi? Does 6,000 spi yield twice the detail of 3,000 spi?

Thanks in advance for all insight.
Best, Keefe.

[jb7]

Thanks Harry-
must give this one a go-

j

Just a quick explanation of the noise removal for the V750 scans.
Since the noise has mainly green and magenta components I use the
a channel in Lab on top of the L channel in hard light mode.
Here you can see the reason why it works:

http://www.sooshee.com/tmp/abney/noise_big.jpg

the a channel stores the green/magenta values of an image (green in dark,
magenta in bright, neutral is in the middle)
Since the noise is per pixel you get an isolated version of the noise without
the luma component, the actual detail of the image.
So you use the reversed version of the chroma noise to fill the "gaps" in the
Lightness channel without affecting the image detail much.

The easiest way to do that is just to copy the L channel and the a into a
new grayscale file, combine them in hard light mode and copy the result
back into the L channel.

[Svitantti]

“is there a point at which you get diminishing returns from a given film? In other words, is there a point at which film simply cannot provide additional detail regardless of the quality of the scanner?”

Depends a lot, do you mean only the lenses resolution by this "detail". I guess this is what Sandy already answered.
However, we often want the scans not only to record the properties of the camera system (the lens), but also something that is considered properties of the film itself.

... and this is what people often forget or ignore when talking about scanning resolution.

The film consists of grain and even the faster films have also smaller grains among the larger ones. It is probably not critical to have the smallest grains of a pushed 400 ASA film scanned, but when making a large print out of a film image, it could be also important to reproduce the real grain as well as possible.
At least I would want to do that rather than alter the image during scanning process because I would just want to get the details recorded by the lens.

[sanking]

There is an excellent article on this subject available as a .pdf document from http://aic.stanford.edu/sg/emg/library/
"Film Grain, Resolution and Fundamental Film Particles "(PDF)
Tim Vitale, April 2007

What most people call film grain is is actually accumulation of grain particles and these accumulations vary a great deal in size, from what we can see with a good scanner or microscope to smaller particles that no scanner could capture. So no matter how much resolution is applied it is impossible to capture the smallest grain particles. Vitale points out in the article that the size of film gain is much larger than the ability of film to resolve detail of a specific size, and from this suggests that film grain and the ability to resolve detail are different properties.

Sandy King

[I]
... and this is what people often forget or ignore when talking about scanning resolution.

The film consists of grain and even the faster films have also smaller grains among the larger ones. It is probably not critical to have the smallest grains of a pushed 400 ASA film scanned, but when making a large print out of a film image, it could be also important to reproduce the real grain as well as possible.
At least I would want to do that rather than alter the image during scanning process because I would just want to get the details recorded by the lens.

[Svitantti]

Oh this is interesting... I agree, you can almost endlessly increase the resolution and still get something new.

Still to some extend I say you also want to scan some of the grain and not only what optics draw on the film. I know the "grain" is actually accumulation of grains, but thats just what people call grain. So we want to scan the largest grain groups to some level to make the texture of the photo look somewhat the same that a wet print would. At least many would.

I dont know how much you would need to make it look good. I guess that depends a lot about film speed and of course print size.

[neil poulsen]

Regardless of which is better, consumer flatbed scanners like the Epson 750 or higher end drum scans, the very large majority of scans will be from consumer models and the very large majority of prints from scans will be 16'x20" or smaller, if not 13"x19" or smaller.

Many of the more experienced people with using consumer scanners, like Kirk, Sandy, and others in this thread, refer to the challenges of getting the best scans from consumer grade scanners.

It would be a large benefit, if one or more of these experienced people could collaborate and produce some sort of tutorial on how this can be accomplished for those who are less experienced at using consumer scanners.

Perhaps a tutorial like this could be posted on this and on other sites.