Flatbed vs Film Scanners

[One4All]

Even though it's a flatbed scanner, my Epson Perfection 4870 Photo's
specs exceed those of my Dimage Scan Multi II (not Pro): 48 bit-depth
vs 36; 3.8 Dmax vs 3.6; and 4800 optical resolution vs 2820.
Quality-wise, is there any reason my 4870 should not replace the Dimage
in all film formats, including 35-mm?

BTW, there's advice out there to scan raw images as archival, which can
be manipulated later in PS. There's other advice that the more the
image can be corrected before the scan, the less deterioration that
will occur in PS because of less need for correction. Comments?

 

[rafeb]

Even though it's a flatbed scanner, my Epson Perfection 4870 Photo's
specs exceed those of my Dimage Scan Multi II (not Pro): 48 bit-depth
vs 36; 3.8 Dmax vs 3.6; and 4800 optical resolution vs 2820.
Quality-wise, is there any reason my 4870 should not replace the Dimage
in all film formats, including 35-mm?

What do your eyes tell you, when you
compare the scans?

(To compare the scans, you'd need to
upsample the Minolta output to match
to Epson's 4800.)


rafe b.
http://www.terrapinphoto.com

 

[-]

Like Rafe said, you need to compare with your own eyes. Lots of
manufacturer's specification claims are "theoretical" at best. Dmax is a
good example. Please post back with your opinion because no doubt there are
a lot of people who would like to know your conclusions after doing a
comparison!

Doug

 

[Bart van der Wolf]

Even though it's a flatbed scanner, my Epson Perfection 4870
Photo's specs exceed those of my Dimage Scan Multi II (not
Pro): 48 bit-depth vs 36; 3.8 Dmax vs 3.6; and 4800 optical
resolution vs 2820.
Quality-wise, is there any reason my 4870 should not replace
the Dimage in all film formats, including 35-mm?

Yes, because specs in sales brochures don't always equal performance
in actual use. Practical results are sometimes much lower quality, and
most flatbed scanners are a prime example, especially when compared
with a dedicated film scanner. For larger filmsizes, the better
flatbeds can be a low cost alternative for a dedicated film scanner,
but that is usually not the case for 35mm film.

BTW, there's advice out there to scan raw images as archival,
which can be manipulated later in PS. There's other advice that
the more the image can be corrected before the scan, the less
deterioration that will occur in PS because of less need for
correction. Comments?

There are few things that can improve the technical image quality of a
scan. Exposure and focus are the most important ones that do, and
better control of those generally leads to better results, also after
post-processing. Most settings in scanner drivers are just for
post-processing, something that possibly is best left for a good
photo-editor, like Photoshop.

Bart

 

[One4All]

Like Rafe said, you need to compare with your own eyes. Lots of
manufacturer's specification claims are "theoretical" at best. Dmax is a
good example.

I doubt I'll see any difference on my desktop Epson 1280 inkjet
printer, which means the lower-cost flatbed wins. Besides, if
manufacturers' claims are theoretical, they apply to desktop film
scanners, as well as flatbeds. This leaves me nowhere.

Please post back with your opinion because no doubt there are
a lot of people who would like to know your conclusions after doing a

comparison!

My Dimage Scan software can't find the Minolta film scanner on my Mac
G5, running OS 10.3.8. The scanner connects to the system via an
Adaptec 29160/29160N SCSI card. The system sees both. I've downloaded
and installed DS Multi ver. 2.0.5, but to no avail. I've filed a help
request with Konica/Minolta & it may take some time (a week?) to hear
from them. If you or anyone has some advice, I'd appreciate it.

Meanwhile, I wonder if any computer or computer graphics magazines have
done a comparison between current film and flatbed scanners.

 

[One4All]

(To compare the scans, you'd need to
upsample the Minolta output to match
to Epson's 4800.)

Well, Rafe, doesn't upsampling itself deteriorate quality?

BTW, you've been very helpful, elsewhere in this group.

 

[One4All]

...specs in sales brochures don't always equal performance
in actual use. Practical results are sometimes much lower quality, and
most flatbed scanners are a prime example, especially when compared
with a dedicated film scanner.

Are you saying specs for flatbeds are more unreliable than those for
film scanners? Do you know this through your own experience? Why are
manufacturers of film scanners more pristine than manufacturers of
flatbeds?

There are few things that can improve the technical image quality of a
scan. Exposure and focus are the most important ones that do, and
better control of those generally leads to better results, also after

post-processing. Most settings in scanner drivers are just for
post-processing, something that possibly is best left for a good
photo-editor, like Photoshop.

So, you're saying scan raw. The scanners I have adjust for exposure and
focus automatically. If I scan raw, I will do so at 48 (16/per
channel)-bit depth. Elsewhere in this group, I've learned that
manipulation in PS degrades an image & the more pixels available
results in less degradation when converted to 8-bit for final output.

So, ignore scanner adjustments re: tonality and color balance, etc.
Scan the raw image and adjust later in PS. Right?

 

[rafe bustin]

Well, Rafe, doesn't upsampling itself deteriorate quality?

BTW, you've been very helpful, elsewhere in this group.

Upsampling isn't lossy. Downsampling is,
but less so if the scanner's overrated
for dpi in the first place.

Consider comparing two scanners, scanner A
at 2000 dpi scanner B at 4000 dpi. There
are at least three ways to do this.


1. Upsample A's output to 4000 dpi, compare
scans at 4000. You're diluting A's output
but not degrading it in any way.

2. Pick some intermediate dpi (say, 3000 in
this case.) Upsample A's output to 3000,
and downsample B's output to 3000. In other
words, split the difference.

3. Downsample B's output to 2000 dpi. If
B's dpi rating is _truly_ 4000, that will
be very unfair to B. OTOH, if B's dpi
rating was highly inflated in the first
place, the loss of information is that
much less.


If done right, I think any one of these
three approaches will probably point to
the same conclusion. Option 1 is what I
use. It's perfectly valid.



rafe b.
http://www.terrapinphoto.com

 

[rafe bustin]

Are you saying specs for flatbeds are more unreliable than those for
film scanners? Do you know this through your own experience? Why are
manufacturers of film scanners more pristine than manufacturers of
flatbeds?

I don't know about pristine.

I can't explain it. But it is a fact,
for each of the five or six scanners
I've owned and five or six more that
I've worked with briefly.

Partly it's marketing. Epson's not
exactly lying with their dpi ratings.

I own a flatbed/film scanner, a very
old design, that is rated at 2500 dpi
and IMHO delivers on that rating.

dpi isn't the same thing as MTF,
which is what matters for film.

The Epson 4870 on a good day matches
a 2500-2700 dpi film scanner.

I've seen some scans from a 4870 that
look more like 2000 dpi. Have you seen
the "Scanner Bakeoff?"

<http://www.jamesphotography.ca/scanner_test.html>


rafe b.
http://www.terrapinphoto.com

 

[Leonhard Pang]

(e-mail address removed) (One4All) wrote in <1112069808.680582.124510
@g14g2000cwa.googlegroups.com>:

Are you saying specs for flatbeds are more unreliable than those for
film scanners? Do you know this through your own experience? Why are
manufacturers of film scanners more pristine than manufacturers of
flatbeds?

It's a theoretical specs you get. But the question is, what do you really
get. The german computer magazin c't tested serveral flatbed scanners and
dedicated filmscanners, and the filmscanners where always closer to the
specs then the flatbed.
The flatbed specs were all way overrated. None of the flatbed came close to
the resolution of a dedicated filmscanners.

Here some of the testresults:
Ok.. ct 12/03 page 137
Epson Perfection 3200 Photo 1180x1200 dpi
Canon CanoScan 9900F 1160x1120 dpi
Microtek ArtixScan 1800F 1280x1360 dpi
Epson Perfection 4870 Photo 1700x1600 dpi

Minolta Dimage Multi Pro 2840x4960 dpi

Nikon Coolscan LS-50ED 3600x3600 dpi
Nikon Coolscan LS-5000ED 3600x3600 dpi

And that's only resolutions. The dmax is a different story.
What you get in the specs is the ability of the A/D chip, but what you
really get is also a different story.

You really need to compare by scanning identical film, and then see, what
the scanners are really capable... or you also might just test, what your
camera and lens are capable to resolve.

-Leonhard

 

[Wilfred]

Upsampling isn't lossy. Downsampling is,
but less so if the scanner's overrated
for dpi in the first place.

Consider comparing two scanners, scanner A
at 2000 dpi scanner B at 4000 dpi. There
are at least three ways to do this.


1. Upsample A's output to 4000 dpi, compare
scans at 4000. You're diluting A's output
but not degrading it in any way.

2. Pick some intermediate dpi (say, 3000 in
this case.) Upsample A's output to 3000,
and downsample B's output to 3000. In other
words, split the difference.

3. Downsample B's output to 2000 dpi. If
B's dpi rating is _truly_ 4000, that will
be very unfair to B. OTOH, if B's dpi
rating was highly inflated in the first
place, the loss of information is that
much less.


If done right, I think any one of these
three approaches will probably point to
the same conclusion. Option 1 is what I
use. It's perfectly valid.

I think it depends on your hypothesis. A fair hypothesis would be that
the flatbed is inferior. Thus, downsampling the flatbed's output would
be a good approach to prove that hypothesis. If the comparison does not
prove the hypothesis you can still try another hypothesis and try one of
the other approaches.
The problem with upsampling is often that it leads to soft images which
are hard to compare.

 

[rafe bustin]

I think it depends on your hypothesis. A fair hypothesis would be that
the flatbed is inferior. Thus, downsampling the flatbed's output would
be a good approach to prove that hypothesis. If the comparison does not
prove the hypothesis you can still try another hypothesis and try one of
the other approaches.
The problem with upsampling is often that it leads to soft images which
are hard to compare.

How else do you compare two filmscanners of
different native resolutions?

If A has half the native resolution of B,
it could still look sharper after upsampling --
if it has a significantly better MTF than B.

Conversely, if B has a sufficiencly poor MTF
compared to A, it might still look worse than
A after downsampling.

If both scanners are offering the max possible
MTF, then it's a no brainer - the scanner with
the higher resolution wins.


rafe b.
http://www.terrapinphoto.com

 

[Bart van der Wolf]

SNIP

Are you saying specs for flatbeds are more unreliable than
those for film scanners?

I'm saying that the DPI (they should specify PPI or SPI) specification
of a flatbed scanner is usualy not a good measure of its real
resolution. You can take as many samples as you wish, if the lens
doesn't resolve the detail, then the higher number of samples does
very little to improve the resolution.
The PPI specs of filmscanners is usually a much closer approximation
of real performance, because in general the lenses are better.

Do you know this through your own experience?

Yes, and from other people reporting actual resolution measurements.
You can Google for posts about a slanted edge target (SET) evaluation.

Why are manufacturers of film scanners more pristine than
manufacturers of flatbeds?

It has more to do with the lack of knowledge of marketoids. Of course
it is easier to sell high spec numbers than a lower resolution number.

SNIP

So, you're saying scan raw.

Either that, or don't waste too much time tweaking things that are
probably better implemented in Photoshop, if that is what you are
going to use anyway.

The scanners I have adjust for exposure and focus
automatically. If I scan raw, I will do so at 48 (16/per
channel)-bit depth.

That's good, and after the major tonemapping and color corrections,
there is not much reason to stay in 48-bits, since allmost all output
modalities are 24-bit devices.

Elsewhere in this group, I've learned that manipulation in PS
degrades an image & the more pixels available results in less
degradation when converted to 8-bit for final output.

It helps to do the major postprocessing in 48-bits and at full native
resolution of the capture device. You can always down-sample for the
final purpose and the loss of information is minimal (for the output
size needed). Upsizing can only add fabricated info.

So, ignore scanner adjustments re: tonality and color balance,
etc. Scan the raw image and adjust later in PS. Right?

That sums it up, yep.

Bart

 

[Wilfred]

How else do you compare two filmscanners of
different native resolutions?

My statement referred to a situation where a flatbed is compared to a
filmscanner with a lower native resolution (the OP was comparing an
4800ppi Epson to a 2820ppi Minolta [A]).

If A has half the native resolution of B,
it could still look sharper after upsampling --
if it has a significantly better MTF than B.

This could also be an approach to prove the hypothesis 'A is better than
B'. However, if A *just a bit* better than B I don't know if you will be
able to see the difference in quality between B's unsharp hi-res scan
ans A's low-res scan that has become blurred through the upsampling
process. OTOH some kind of smart upsampling algorithm (e.g., Genuine
Fractals or one of those algorithms evaluated on Bart van der Wolf's
site) might still be able to show the difference. But there will be some
uncertainty as to whether you will be able to see the difference. It is
for that reason that I would favor the other approach:

Conversely, if B has a sufficiencly poor MTF
compared to A, it might still look worse than
A after downsampling.

This was the approach I was referring to. I think in that case (or even
if B's MTF is *just a bit* worse than A's) it will be impossible to make
B's image look sharper then A's unless you play dirty tricks. Thus, the
hypothesis 'A is better than B' is easily proven.

If both scanners are offering the max possible
MTF, then it's a no brainer - the scanner with
the higher resolution wins.

Yes but if you know the MTF in advance you don't need to carry out this
simple up-/downsampling experiment where the judgment depends on human
perception;-)

 

[Hecate]

How else do you compare two filmscanners of
different native resolutions?

Look at the DMax which is what is really important. The Epson has been
found to have a lower DMax than that claimed by Epson and if you check
Epson's literature you will see that they claim this DMax under only
one specific set of conditions. The DMax in film scanners, whilst
still overrated, is much closer to the theoretical maximum quoted by
the manufacture. Whilst the pixel count is important in terms of image
size, the DMax is far more important in terms of image quality.

--

Hecate - The Real One
(e-mail address removed)
Fashion: Buying things you don't need, with money
you don't have, to impress people you don't like...

 

[Hecate]

Are you saying specs for flatbeds are more unreliable than those for
film scanners?

I assume he is, and he's right. It has to do with the theoretical DMax
which is often higher with dedicated film scanners and is much closer
to theoretical than that of flatbeds. If you search Google Groups
you'll find discussions on the theoretical as opposed to actual DM<ax
of flatbeds and film scanners and see that the film scanners Real
World DMax is much closer as I said. In actual fact the DMax has far
more bearing on the quality of the scanned image than the number of
pixels available. No doubt you will find the same info searching the
Net if you care to.

Do you know this through your own experience? Why are
manufacturers of film scanners more pristine than manufacturers of
flatbeds?

No, it's just that the capabilities of dedicated film scanners,
regardless of manufacturer are much higher.

--

Hecate - The Real One
(e-mail address removed)
Fashion: Buying things you don't need, with money
you don't have, to impress people you don't like...

 

[Leonhard Pang]

(e-mail address removed) (One4All) wrote in <1111865395.466888.116000
@f14g2000cwb.googlegroups.com>:

Check:
http://www.heise.de/ct/ftp/testbilder/scanner_0412/

especially the USAF-Testchart at the max resolutions you see the
differences.

-Leonhard

 

[Ole-Hjalmar Kristensen]

<snip>

BvdW> There are few things that can improve the technical image quality of a
BvdW> scan. Exposure and focus are the most important ones that do, and
BvdW> better control of those generally leads to better results, also after
BvdW> post-processing. Most settings in scanner drivers are just for
BvdW> post-processing, something that possibly is best left for a good
BvdW> photo-editor, like Photoshop.


BvdW> Bart

Yes I agree. Interestingly, even on my Epson 4990, a pice of 35mm film can
curl enough that it is clearly out of focus (easily verified by
turning it upside down and scanning again).

 

[ian lincoln]

I don't know about pristine.

I can't explain it. But it is a fact,
for each of the five or six scanners
I've owned and five or six more that
I've worked with briefly.

Partly it's marketing. Epson's not
exactly lying with their dpi ratings.

I own a flatbed/film scanner, a very
old design, that is rated at 2500 dpi
and IMHO delivers on that rating.

dpi isn't the same thing as MTF,
which is what matters for film.

The Epson 4870 on a good day matches
a 2500-2700 dpi film scanner.

I've seen some scans from a 4870 that
look more like 2000 dpi. Have you seen
the "Scanner Bakeoff?"

<http://www.jamesphotography.ca/scanner_test.html>

i'd like to see these results. i have the scan elite II.It is incredibly
slow, crashes and has incredible levels of noise on thin slides and all
negatives. Was going to give up and get the canon 8400 or the 9000F

 

[Anoni Moose]

My Konica-Minolta 5400 film scanner has beautifully clean noise-free
blacks (on slide film). SO much better than my flatbed film-scanner
(old epson photo 2450 (or was it 2540?)) or my even older Polaroid film
scanner. No crashes, absolutely beautiful scans, and its 2x2
carriage is easily adaptable to RBT Stereo mounts (for those who know
what those are).

Mike