Film Scanner DPI vs DSLR Megapixels

Hello again.

Just at my final step for the purchase of the Minolta Dimage Scan Elite
II, I need some clarification on resolution values: DPI vs Megapixels.
The Elite II is a 2820 DPI scanner, which I have calculated to be equal
to a 10 MP full-frame (35mm) DSLR. My calculation was:
Short edge of 35mm film: 24 / 25.4 = 0.945"
Long edge of 35mm film: 36 / 25.4 = 1.417"
Pixel count along the short edge = 0.945 * 2820 = 2664
Pixel count along the long edge = 1.417 * 2820 = 3995
Total pixel count on the scanning area = 2664 * 3995 = 10642680,
roughly 10M

Now, should this calculation mean that a good slide + scanning with
this scanner will be equivalent to shooting with a 10MP full-frame 35mm
DSLR ? How would the results compare to the line of current 8MP
not-full-frame DSLRs and would results be comparable to those off an
EOS 5D (which has a 12.8 MP 35mm full-frame sensor) by any criteria?

Best,
-arifi

"arifi" schreef in bericht
ps.com...
Hello again.

Just at my final step for the purchase of the Minolta Dimage Scan Elite
II, I need some clarification on resolution values: DPI vs Megapixels.
The Elite II is a 2820 DPI scanner, which I have calculated to be equal
to a 10 MP full-frame (35mm) DSLR. My calculation was:
Short edge of 35mm film: 24 / 25.4 = 0.945"
Long edge of 35mm film: 36 / 25.4 = 1.417"
Pixel count along the short edge = 0.945 * 2820 = 2664
Pixel count along the long edge = 1.417 * 2820 = 3995
Total pixel count on the scanning area = 2664 * 3995 = 10642680,
roughly 10M

Suppose that your Minolta Dimage Scan Elite produces within your
calculation perfect pixels. (They are not).
Then you have 10 M pixels for each of the three colors. That works
out to be 10 M for R, 10 M for G and 10 M for B. This is 30 M
subpixels.

The pixel count in a camera starts of with subpixels.
1/2 the pixels are green, 1/4 are Blue, 1/4 are Red.
So to get the same number of perfect pixels you would need 3 times
(or 4 times) as many pixels in the DSLR.
And then the DSLR still can produce more moire on structures because
the space between the different colors is devided.
(Only one 1/4 of the area is sensitive to red for example, where with the
scanner
this is often near 100 %, this depends on the design though).

So the technical design of the scanner could produce far better results than
the DSLR. But there are a lot of limits for the scanner, film or slides in
practise do not have the resolution that the scanner is specified for.
The scanner probably is not totaly sharp. With desktop scanners I have
scanned raisor blades to see how sharp the scanners could scan this,
maybe you could come up with a setup to try this with your scanner and
probably you'll see that the edge of the totaly sharp raisor blades covers
several pixels in width.

So if you had totaly sharp slides (which do not exist) and a totaly perfect
scanner with your specifications. You would need more than 30 Mp to
compete with that setup.

ben

Now, should this calculation mean that a good slide + scanning with
this scanner will be equivalent to shooting with a 10MP full-frame 35mm
DSLR ? How would the results compare to the line of current 8MP
not-full-frame DSLRs and would results be comparable to those off an
EOS 5D (which has a 12.8 MP 35mm full-frame sensor) by any criteria?

Best,
-arifi

"arifi" wrote:
Hello again.

Just at my final step for the purchase of the Minolta Dimage Scan Elite
II, I need some clarification on resolution values: DPI vs Megapixels.
The Elite II is a 2820 DPI scanner, which I have calculated to be equal
to a 10 MP full-frame (35mm) DSLR. My calculation was:

People will argue, but in real life 35 film _no matter how you scan or print
it_ at its very very best captures about 8MP of detail. (Roger Clark, one of
the more sensible people here, still has a rather strange appreciation for
Velvia 50, which in my experience it's grainier and uglier than Provia 100F,
and argues that it's "worth" a lot more MP than I've ever seen from a square
mm of film (and I've spent the last four years shooting medium format film
while waiting for the 5D, inspecting every frame with a 60x microscope.))

Anyway, in real life, 35mm makes a _nice_ 8x10, but is losing it badly
compared to medium format at 11x14.

Short edge of 35mm film: 24 / 25.4 = 0.945"
Long edge of 35mm film: 36 / 25.4 = 1.417"
Pixel count along the short edge = 0.945 * 2820 = 2664
Pixel count along the long edge = 1.417 * 2820 = 3995
Total pixel count on the scanning area = 2664 * 3995 = 10642680,
roughly 10M

The problem here is that scanned pixels are really really ugly. Incredibly
ugly. Here's a page of some of the best scans made by modern equipment.

http://www.terrapinphoto.com/jmdavis/

Now, should this calculation mean that a good slide + scanning with
this scanner will be equivalent to shooting with a 10MP full-frame 35mm
DSLR ? How would the results compare to the line of current 8MP
not-full-frame DSLRs and would results be comparable to those off an
EOS 5D (which has a 12.8 MP 35mm full-frame sensor) by any criteria?

The 5D competes with 645 (56 x 42 mm); 35mm isn't even close.

David J. Littleboy
Tokyo, Japan

arifi,
MP is convenient when comparing cameras, but with files what counts is
the uncompressed size.With your scanner that will be approx 30mb file.
Close to a 10MP camera, But no scanner is perfect, where as your lens -
sensor alignment on a camera is close to perfect. Even a well scanned
35mm slide will not equal a camera file in sharpness but it should be
better than the camera file in dynamic range. That is why folk just
don't copy slides with a DSLR. I'm not going to make what equals what
pronouncements. A couple of years ago I was with a group photographing
a AAA minor league baseball team. We had to have a team photo for
opening day, 4 days hence, cheap guys so they wanted to run their years
supply all at once. We shot a Mamiya 645 and a 6mp Kodak 760, one of
the owners friends used a Hassleblad. Because of the printers time
constraints they used our 6mp image. Nice team pic, better than the
year before's, we don't know how that was shot. How did the images
compare, the team had enlargements made from a Mamiya and Hassy image
to the 20x24 size of the printed team picture. Our digital file printed
was way better than the Mamiya file (admittedly an old camera ald
lens), the enlarged (photographic print) from the Hassleblad was
marginally better than the offset printed digital. The digital file
also kept the team to their deadlines, no extra rush fees. Convinced
me.
A scanner is the best way to reproduce your 35mm slides & negs, can
they be as good as digital camera files, probably not.

Tom

David J. Littleboy wrote:
"arifi" wrote:

Hello again.

Just at my final step for the purchase of the Minolta Dimage Scan Elite
II, I need some clarification on resolution values: DPI vs Megapixels.
The Elite II is a 2820 DPI scanner, which I have calculated to be equal
to a 10 MP full-frame (35mm) DSLR. My calculation was:


People will argue, but in real life 35 film _no matter how you scan or print
it_ at its very very best captures about 8MP of detail. (Roger Clark, one of
the more sensible people here, still has a rather strange appreciation for
Velvia 50, which in my experience it's grainier and uglier than Provia 100F,
and argues that it's "worth" a lot more MP than I've ever seen from a square
mm of film (and I've spent the last four years shooting medium format film
while waiting for the 5D, inspecting every frame with a 60x microscope.))

David,
The technical specifications of velvia show it has higher spatial
resolution than provia. Then you mis-characterize my position.
Don't confuse spatial resolution with image quality. Fine
grained 35mm film has higher spatial resolution than most DSLRs, but
film also has much poorer signal-to-noise ratios. Try reading:
http://www.clarkvision.com/imagedeta....summary1.html
and get beyond Figure 1 (digital megapixel equivalent versus film speed);
read the Apparent Image Quality section. There you will see I
rate an 8-megapixel DSLR well above both velvia and provia 35mm film.

For the OP: see the above page plus:
Image Detail (How much detail can you capture and scan?)
http://www.clarkvision.com/imagedetail/scandetail.html

But note, as above, spatial resolution is only one part of
image quality.

Also note that the reasonable rating of a scanner ppi is about 2/3
the manufacturer's specification, so 2820 ppi consumer scanner
would be equivalent to about 1890 ppi compared to a high end
scanner like a drum scan.

Roger

"arifi" writes:

Just at my final step for the purchase of the Minolta Dimage Scan Elite
II, I need some clarification on resolution values: DPI vs Megapixels.
The Elite II is a 2820 DPI scanner, which I have calculated to be equal
to a 10 MP full-frame (35mm) DSLR. My calculation was:
Short edge of 35mm film: 24 / 25.4 = 0.945"
Long edge of 35mm film: 36 / 25.4 = 1.417"
Pixel count along the short edge = 0.945 * 2820 = 2664
Pixel count along the long edge = 1.417 * 2820 = 3995
Total pixel count on the scanning area = 2664 * 3995 = 10642680,
roughly 10M

Now, should this calculation mean that a good slide + scanning with
this scanner will be equivalent to shooting with a 10MP full-frame 35mm
DSLR ? How would the results compare to the line of current 8MP
not-full-frame DSLRs and would results be comparable to those off an
EOS 5D (which has a 12.8 MP 35mm full-frame sensor) by any criteria?

Digital original pixels are 'worth more' in image quality than scanned
pixels -- they're one generation closer to the original, after all.
--
David Dyer-Bennet, , http://www.dd-b.net/dd-b/
RKBA: http://www.dd-b.net/carry/
Pics: http://dd-b.lighthunters.net/ http://www.dd-b.net/dd-b/SnapshotAlbum/
Dragaera/Steven Brust: http://dragaera.info/

I am not sure of the technical aspects, however I have recently started
scanning my 35mm negatives using a Nikon 4000DPI scanner.

I would never be able to use the 100% scan to print, as it just isn't sharp
enough (in fact I thought the scans would be sharper than they were), but I
would be able to print a 100% photo from my 20D.

With the scan, you need to downsize the image to get a decent image.

I suppose it is similar to comparing digital camera's by the amount of
megapixels. You can have a cheap 6MP point and shoot camera, and an
expensive 6MP camera with a decent lens. They are both 6MP, but what is the
point of having a sensor that can capture 6MP, when the image reaching the
sensor is crap. If you take a photo on a DSLR with no lens attached, the
image is still 6MP, but just a big blur.


"arifi" wrote in message
ps.com...
Hello again.

Just at my final step for the purchase of the Minolta Dimage Scan Elite
II, I need some clarification on resolution values: DPI vs Megapixels.
The Elite II is a 2820 DPI scanner, which I have calculated to be equal
to a 10 MP full-frame (35mm) DSLR. My calculation was:
Short edge of 35mm film: 24 / 25.4 = 0.945"
Long edge of 35mm film: 36 / 25.4 = 1.417"
Pixel count along the short edge = 0.945 * 2820 = 2664
Pixel count along the long edge = 1.417 * 2820 = 3995
Total pixel count on the scanning area = 2664 * 3995 = 10642680,
roughly 10M

Now, should this calculation mean that a good slide + scanning with
this scanner will be equivalent to shooting with a 10MP full-frame 35mm
DSLR ? How would the results compare to the line of current 8MP
not-full-frame DSLRs and would results be comparable to those off an
EOS 5D (which has a 12.8 MP 35mm full-frame sensor) by any criteria?

Best,
-arifi

"Roger N. Clark (change username to rnclark)" wrote:
David J. Littleboy wrote:
"arifi" wrote:

Hello again.

Just at my final step for the purchase of the Minolta Dimage Scan Elite
II, I need some clarification on resolution values: DPI vs. Megapixels.
The Elite II is a 2820 DPI scanner, which I have calculated to be equal
to a 10 MP full-frame (35mm) DSLR. My calculation was:


People will argue, but in real life 35 film _no matter how you scan or
print it_ at its very very best captures about 8MP of detail. (Roger
Clark, one of the more sensible people here, still has a rather strange
appreciation for Velvia 50, which in my experience it's grainier and
uglier than Provia 100F, and argues that it's "worth" a lot more MP than
I've ever seen from a square mm of film (and I've spent the last four
years shooting medium format film while waiting for the 5D, inspecting
every frame with a 60x microscope.))

David,
The technical specifications of velvia show it has higher spatial
resolution than provia.

Maybe. But the experience here is that Velvia 50 scans uglier than the
"100F" films.

Then you mis-characterize my position.
Don't confuse spatial resolution with image quality. Fine
grained 35mm film has higher spatial resolution than most DSLRs, but
film also has much poorer signal-to-noise ratios. Try reading:
http://www.clarkvision.com/imagedeta....summary1.html
and get beyond Figure 1 (digital megapixel equivalent versus film speed);

Oops: you've redone that figure since the last time I looked. Still, I think
you are way overstating 35mm Velvia. It really takes 6x7 before you see more
detail from film than from 12MP digital.

read the Apparent Image Quality section. There you will see I
rate an 8-megapixel DSLR well above both velvia and provia 35mm film.

Well, if you didn't start out by calling it at 14MP, then I wouldn't grump.

The problem is that film has a long "tail" in its response; it can record
extremely high contrast detail way out to insane lp/mm levels, but does very
badly at lower contrast images. In my (and others) tests of 6x7 Provia vs.
the 12MP cameras, 6x7 nails the fine high-contrast detail that the 5D is
just beginning to lose it on, but the rest of the image is a toss-up. I've
found that for portraits, it is incredibly critical to fill the frame with
the subject with film. Standing back a bit to take in some of the context is
simply not an option. Because film is so grody for low-contrast parts of the
image.

In other words, Velvia/Provia are only "14MP" for street signs and license
plates. For everything else, they're not even 8MP.

Also note that the reasonable rating of a scanner ppi is about 2/3
the manufacturer's specification, so 2820 ppi consumer scanner
would be equivalent to about 1890 ppi compared to a high end
scanner like a drum scan.

I suspect that this rule of thumb, while in the right direction, is a tad
too general. I'd guess the 2820 ppi class scanners' practical/apparent
resolutions are a lot closer to their nominal values than the 4000 ppi class
scanners. This has to be true simply because film begins to break down at
anything over an 8x enlargement, so no matter what resolution you scan at,
2400 ppi is about the highest resolution film itself supports _for producing
grain-sniffable prints_. Posters on a wall is a different question, of
course.

David J. Littleboy
Tokyo, Japan

David J. Littleboy wrote:

"Roger N. Clark (change username to rnclark)" wrote:
The technical specifications of velvia show it has higher spatial
resolution than provia.

Maybe. But the experience here is that Velvia 50 scans uglier than the
"100F" films.

Maybe it is your scanner. My experience is just the opposite.

Then you mis-characterize my position.
Don't confuse spatial resolution with image quality. Fine
grained 35mm film has higher spatial resolution than most DSLRs, but
film also has much poorer signal-to-noise ratios. Try reading:
http://www.clarkvision.com/imagedeta....summary1.html
and get beyond Figure 1 (digital megapixel equivalent versus film speed);

Oops: you've redone that figure since the last time I looked. Still, I think
you are way overstating 35mm Velvia. It really takes 6x7 before you see more
detail from film than from 12MP digital.

Remember the Figure 1 plot is only spatial resolution, and breaks
down into two components: luminance resolution and color resolution.
The bottom of the vertical bar for each film is the luminance
megapixel equivalent and for Provia 100, that is 8 megapixels.
But bayer sensors have lower color resolution, so your quoting
the 12 megapixel number is for color resolution. Most of the
effect that you see in a typical image is luminance resolution.

Then, second, again it depends on your scanner. I used multiple
scanners in my research including high-end drum scanners.

If you are simply comparing image quality, the higher
signal-to-noise of electronic sensors further improves image quality.
It takes some effort to separate these effects, and that is what
I attempted to do in the article.

read the Apparent Image Quality section. There you will see I
rate an 8-megapixel DSLR well above both velvia and provia 35mm film.

Well, if you didn't start out by calling it at 14MP, then I wouldn't grump.

The problem is that film has a long "tail" in its response; it can record
extremely high contrast detail way out to insane lp/mm levels, but does very
badly at lower contrast images. In my (and others) tests of 6x7 Provia vs.
the 12MP cameras, 6x7 nails the fine high-contrast detail that the 5D is
just beginning to lose it on, but the rest of the image is a toss-up. I've
found that for portraits, it is incredibly critical to fill the frame with
the subject with film. Standing back a bit to take in some of the context is
simply not an option. Because film is so grody for low-contrast parts of the
image.

Again you seem focused on Figure 1. Try focusing on the table in the
AIQ section. Perhaps I need to make that data into a picture. I'm slowly
learning that photographers can't seem to comprehend numbers; they
need a picture ;-).

In other words, Velvia/Provia are only "14MP" for street signs and license
plates. For everything else, they're not even 8MP.

And if you looked at the AIQ table, you would see I agree:
35mm Velvia 50 has an AIQ = 14; 8 MPixel DSLR = 38 at ISO 100,
over twice as good.


Also note that the reasonable rating of a scanner ppi is about 2/3
the manufacturer's specification, so 2820 ppi consumer scanner
would be equivalent to about 1890 ppi compared to a high end
scanner like a drum scan.

I suspect that this rule of thumb, while in the right direction, is a tad
too general. I'd guess the 2820 ppi class scanners' practical/apparent
resolutions are a lot closer to their nominal values than the 4000 ppi class
scanners.

I have yet to see a consumer scanner that did better than the 2/3 guideline.
The low ppi scanners tend to be made more cheaply, so all I've seen
seem the follow the guide. The manufacturers should be required
to publish the MTF for their rated PPI.

This has to be true simply because film begins to break down at
anything over an 8x enlargement, so no matter what resolution you scan at,
2400 ppi is about the highest resolution film itself supports _for producing
grain-sniffable prints_. Posters on a wall is a different question, of
course.

I disagree. I have made 19x24 and 24x36 inch prints (now hanging in
offices and galleries) from 35 mm velvia drum-scanned images that
pro photographers have asked: "is that medium or large format?"
Even at 4000 ppi velvia, Provia and Kodachrome are grain-aliased
by the scanner. Velvia and Kodachrome even at 5000 ppi. This means
that grain appears enhanced at these ppi. But as one goes higher,
the grain aliasing reduces, apparent grain becomes smaller and
the image smoother.

I've seen 30x40-inch prints of 35mm Kodachrome 25 drum scanned at
around 9000 ppi and printed on a Lightjet that I would have
sworn was large format (my mouth dropped to the floor).
Quality can be had from film, but it comes only with top
equipment and a high price. I now only do film in 4x5 now,
otherwise I'm all digital, and looking to replace most 4x5 with
digital mosaics.

Roger
Photos at: http://www.clarkvision.com

"Roger N. Clark (change username to rnclark)" wrote:

Maybe. But the experience here is that Velvia 50 scans uglier than the
"100F" films.

Maybe it is your scanner. My experience is just the opposite.

I compare what the scanner gets with what a (cheap) micropscope and an
expensive high-power loupe show, and don't see a lot of difference. Velvia
50 is relatively grainy stuff compared to the 100F films. If there's detail
captured, it's so hidden in the grain that it's not photographically
significant.

But we've been through this befo Provia 100F is soft at 13x, but its not
ugly, and your graphs of Velvia 50 noise overstate the problem with film
noise.

Oops: you've redone that figure since the last time I looked. Still, I
think
you are way overstating 35mm Velvia. It really takes 6x7 before you see
more detail from film than from 12MP digital.

Remember the Figure 1 plot is only spatial resolution, and breaks
down into two components: luminance resolution and color resolution.
The bottom of the vertical bar for each film is the luminance
megapixel equivalent and for Provia 100, that is 8 megapixels.
But bayer sensors have lower color resolution, so your quoting
the 12 megapixel number is for color resolution. Most of the
effect that you see in a typical image is luminance resolution.

Well, yes. The human eye's color resolution is a tiny fraction of its
luminance resolution, so you are reporting effects that can't be seen...

Then, second, again it depends on your scanner. I used multiple
scanners in my research including high-end drum scanners.

I've not seen anything anywhere near a factor of two difference in
information captured between 4000 ppi Nikon scanners and drum scanners. Of
course, the drum scanner fans do, but it looks to me that they're
halucinating.

read the Apparent Image Quality section. There you will see I
rate an 8-megapixel DSLR well above both velvia and provia 35mm film.

Well, if you didn't start out by calling it at 14MP, then I wouldn't
grump.

The problem is that film has a long "tail" in its response; it can record
extremely high contrast detail way out to insane lp/mm levels, but does
very
badly at lower contrast images. In my (and others) tests of 6x7 Provia
vs.
the 12MP cameras, 6x7 nails the fine high-contrast detail that the 5D is
just beginning to lose it on, but the rest of the image is a toss-up.
I've found that for portraits, it is incredibly critical to fill the
frame with the subject with film. Standing back a bit to take in some of
the context is simply not an option. Because film is so grody for
low-contrast parts of the image.

Again you seem focused on Figure 1.

It's the first thing that appears in the article. It's hard not tog.

Try focusing on the table in the
AIQ section. Perhaps I need to make that data into a picture. I'm slowly
learning that photographers can't seem to comprehend numbers; they
need a picture ;-).

In other words, Velvia/Provia are only "14MP" for street signs and
license
plates. For everything else, they're not even 8MP.

And if you looked at the AIQ table, you would see I agree:
35mm Velvia 50 has an AIQ = 14; 8 MPixel DSLR = 38 at ISO 100,
over twice as good.

That's more reasonable.

SNIP

I don't like not being able to walk up to a print and see more detail, and
your examples of higher resolution scans didn't convince me that more
information was coming off the film at higher resolutions; sure, the grain
structure was being resolved better but I thought the claims for more
subject detail were problematic. Again, all the experience here is that
13x19 from 6x7 is way better than 11x14 from 35mm. So talking about 24x36
prints from 35mm sounds rather odd to me.

David J. Littleboy
Tokyo, Japan