DMAX Question

[Philip Homburg]

At the dawn of digital imaging, I thought that digital cameras and
scanners could do exactly what you described, and solve these problems
faced by wet imaging users. Capturing and merging multiple images to
result in variable dynamic range or focus can be done on the digital
cameras and scanners themselves. After all it is only firmware
manipulation that requires no (or little) additional hardware. Of
course, it takes longer to create such an image, and also takes more
storage. But as of today, no manufacturers are even attempting at these
features. They are all left to the users like Don to post process after
capturing.

The problem here is that the hardware manufactures want to be in the software
business. The D2X encrypted white-balance distaster is the latest example of
this.

Simply documenting the hardware, like many companies used to do (certainly
for professional products) is old fashioned. These days it is "no user
servicable parts inside".

The firmware is limited to the creativity of the hardware designed, and
unfortunately, hardware designers are not very creative when it comes to
designing software.

Of course, the advantage (for the manufacturer) of not documenting the
hardware is that you can sell almost the same hardware for significantly
different prices, depending on the firmware (or just another wire).

 

[Wilfred]

What would be interesting (but you can't do that unless you have access
to the firmware sources), it to implement some kind of multi-sample scanning
but with different exposure times.

This is what VueScan tries to do if you set it to 'long exposure pass'.
The problem is that it can only do this in two separate passes, and that
VueScan isn't very good at combining the two resulting images -
strange artifacts appear in crisp transitions between dark and light areas.

Indeed it would be nice to have a scanner capable of doing single-pass
multi-scanning with different exposures. Technically it shouldn't be too
difficult. It wouldn't surprise me if the marketeers at Nikon and Konica
Minolta are consciously delaying the introduction of this feature, just
to be able to introduce this improvement in some future model when
further increasing ppi & Dmax isn't possible, or -from a marketing point
of view- doesn't make sense anymore.

 

[Alan Smithee]

This is what VueScan tries to do if you set it to 'long exposure
pass'. The problem is that it can only do this in two separate
passes, and that VueScan isn't very good at combining the two
resulting images -
strange artifacts appear in crisp transitions between dark and light
areas.

I've tried the long exposure pass function. I had bad results when I used
it, erroneously, on a negative which didn't require it. I've found it does a
good job on slides which are too dark.

 

[Don]

Give me a day to go back over my old posts and clean it all up into a
single (somewhat) presentable post. I was meaning to do this before
but I'm just perennially busy...

OK, have a look at the "Twin scan" thread...

But first, make sure you get a beverage and a comfy chair... ;o) It's
loooong!

Don.

 

[Don]

Indeed it would be nice to have a scanner capable of doing single-pass
multi-scanning with different exposures. Technically it shouldn't be too
difficult. It wouldn't surprise me if the marketeers at Nikon and Konica
Minolta are consciously delaying the introduction of this feature, just
to be able to introduce this improvement in some future model when
further increasing ppi & Dmax isn't possible, or -from a marketing point
of view- doesn't make sense anymore.

That's it exactly! If they implemented "dynamic exposure" (and,
indeed, "dynamic focus") it would - at a single stroke - eliminate the
need for multiscanning and the race to ever more bits of dynamic
range.

So, it's clearly a marketing issue.

For example, there's no technical reason for LS-50 not to have
multiscanning. But Nikon went through great effort to make enabling it
extremely difficult, exactly so they can overcharge for the LS-5000.

How does that saying go: "First, we kill all the lawyers...". Well,
add marketroids to that. ;o)

Don.

 

[Don]

Simply documenting the hardware, like many companies used to do (certainly
for professional products) is old fashioned. These days it is "no user
servicable parts inside".

I remember the days when it was considered an absolute minimum to
include the schematics with the hardware.

The firmware is limited to the creativity of the hardware designed, and
unfortunately, hardware designers are not very creative when it comes to
designing software.

I don't think it's the engineers. They are clearly being muzzled by
marketroids.

Of course, the advantage (for the manufacturer) of not documenting the
hardware is that you can sell almost the same hardware for significantly
different prices, depending on the firmware (or just another wire).

Exactly! It's all down to marketing.

Don.

 

[Don]

Thanks. By the way, if any of the steps are specific to the Windows
version of NikonScan, you might mention that since I am using the Mac
version.

No, it's not software specific. It's just a method, as you can see
from the parallel post. You could even use VueScan - although I
wouldn't recommend that for other reasons... ;o)

Also, there are differences between versions of NikonScan (e.g., v3 vs
v4) that might be important in your discussion. I use v3.1.2 with the
LS2000 and scan...you guessed it...Kodachromes.

Oh boy... Those Kodachromes are going to be the end of me... Two years
and counting...

I wrestled with them on my LS-30 which doesn't even have Kodachrome
mode and it was murder. BTW, does LS2000 have the Kodachrome mode?

Actually, that's what I'm trying to tackle right now. I find that the
Kodachrome mode on the LS-50 just doesn't go far enough. Especially,
when having to boost the exposure considerably, like > 1 or 2 AG.

So, I've now come back full circle to try and establish what does the
Kodachrome mode (algorithm) actually do?

This is proving to be yet another nightmare as I can't seem to find
any rhyme or reason. It's clear that it boosts red considerably and
green very slightly, while leaving blue untouched. That much I knew
from my LS-30 days. The problem is figuring out *exactly* the amount
of analog gain applied to red and green. And that seems to be all over
the place...

I can approximate the values, but that's not good enough, especially
as the ratio appears to change with exposure (at least in my initial
tests).

However, I would really like reliable values so I can streamline the
process instead of agonizing over each slide individually.

Don.

 

[gazza95]

Hi Kennedy

Thanks for the informative reply. From the slides I scanned there was
clipping at both ends of the scale. So it was clear I was always going
to struggle to get goods scans. Also from a subjective point I found
that any noise in the dark areas was very noticeable, while the noise
in the light areas when scanning negatives is a lot less noticable. Yes
your right in that the negatives do suffer from noise across wider
density ranges.

If I ever replace the Canon fs2710 then I would look at scanning slides
again and be a little more carefull with exposures.

 

[David Blanchard]

Oh boy... Those Kodachromes are going to be the end of me... Two years
and counting...

I wrestled with them on my LS-30 which doesn't even have Kodachrome
mode and it was murder. BTW, does LS2000 have the Kodachrome mode?

Nope. Wish it did.

-db-

 

[Ole-Hjalmar Kristensen]

I have been reading all his reviews of the Epson flatbeds, and my
general impression is that he tends to be a bit optimistic about the
capabilities of these scanners, and that his example images which
show how detailed scans you can get are so small it is difficult to
get an impression of the overall effects of the heavy sharpening on
the image.


AS> OK say he IS talking about slides. What effect is this going to have on the
AS> "highlight" areas of a slide. It seems the statement should be about one
AS> type of film or the other and make a broad statement like that is wrong or
AS> willfully ignorant on the reviewer's part.
AS> Link to review:
AS> http://www.photo-i.co.uk/Reviews/interactive/Epson 4990/Page 1.htm

 

[Don]

Nope. Wish it did.

In that case, boost red Analog Gain (a lot) and green (a little). To
get a feel for it use an area which should be white (or neutral) and
keep adjusting red and green Analog Gain until it becomes white (or
neutral).

Don't worry about any residue cast, because the purpose of AG boost is
not to remove the cast (you'll do that later in your image editor) but
to boost red sufficiently to give you enough range to perform all
those edits later without causing banding.

BTW, have you seen the "Twin Scan" post? Does that help any? (It's a
lot to digest, I know...)

Don.

 

[David Blanchard]

BTW, have you seen the "Twin Scan" post? Does that help any? (It's a
lot to digest, I know...)

Don.

No...that thread has not appeared on my news server. I did manage to
find it in Google groups. (I just recently changed news servers and
was concerned that some threads appeared to be missing. Now I know
for certain that is happening.)

David

 

[Don]

No...that thread has not appeared on my news server. I did manage to
find it in Google groups. (I just recently changed news servers and
was concerned that some threads appeared to be missing. Now I know
for certain that is happening.)

Maybe it's because of the message size? Although ~250 lines is not all
that big.

Did my message pointing to it get through? (the one with: 'OK, have a
look at the "Twin scan" thread...')

But missing posts are very frustrating. For example, you were waiting
for the post and I was waiting for the comments. It's good that I
asked otherwise we both would've had a very long wait... ;o)

BTW, I had another idea about doing Kodachromes using only regular
Positive setting. I'm planning to turn off the Kodachrome setting
because it just doesn't go far enough. So this may be relevant to you
too. I should know more in a day or two and if this new approach works
I'll post a note here.

Don.

 

[David Blanchard]

No...that thread has not appeared on my news server. I did manage to
find it in Google groups. (I just recently changed news servers and
was concerned that some threads appeared to be missing. Now I know
for certain that is happening.)

Maybe it's because of the message size? Although ~250 lines is not all
that big.[/QUOTE]

I don't think that 250 lines is big enough to matter.

Did my message pointing to it get through? (the one with: 'OK, have a
look at the "Twin scan" thread...')

Yes...that one showed up. So I spent a lot of time looking for the
thread until I concluded it was not available on my news server.

BTW, I had another idea about doing Kodachromes using only regular
Positive setting. I'm planning to turn off the Kodachrome setting
because it just doesn't go far enough. So this may be relevant to you
too. I should know more in a day or two and if this new approach works
I'll post a note here.

Thanks...I'll be watching for it.

-db-

 

[Hecate]

That is the general idea - but it doesn't always work as well as
intended. The best-laid schemes o' mice an' men gang aft agley, as one
of my old locals once wrote.

With a name like that you had to be Scottish

--

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...

 

[Kennedy McEwen]

With a name like that you had to be Scottish

That doesn't necessarily follow - the only other guy I have ever met
with the same name was an Australian immigration officer, who held me up
for an hour at Sydney airport about 15years ago to tell me *his* entire
family history, convinced that we must be long lost relatives. We
probably were, but after 27hrs in the air with only a couple of 1hr
stops, I was a little less interested than he was. :-(

But, indeed, I went to the same parish school as the baird himself.

 

[Pavel Dvorak]

No, it wouldn't. The so-called "long exposure pass" just blurs the
image.

Are you sure that this is not a problem of the hardware? My otherwise
miserable CanoScan FS2710 does not show any noticeable bluring with the
long exposure pass, or with other multiple pass selection (long exposure
involves a slow pass plus whatever number of passes you specify, at least
on my scanner; it cannot be run just as a single slow exposure).

There are other problems with that scanner, but they are really a matter
of its design and sensor limitations.

Pavel

 

[Pavel Dvorak]

This is what VueScan tries to do if you set it to 'long exposure pass'.
The problem is that it can only do this in two separate passes, and that
VueScan isn't very good at combining the two resulting images -
strange artifacts appear in crisp transitions between dark and light areas.

I think this is a hardware problem. Does it happen with all scanners? It
does happen with my FS2710, but this shows on a single pass and with the Canon
software as well.

Pavel

 

[Pavel Dvorak]

Hi Kennedy

Thanks for the informative reply. From the slides I scanned there was
clipping at both ends of the scale. So it was clear I was always going
to struggle to get goods scans. Also from a subjective point I found
that any noise in the dark areas was very noticeable, while the noise
in the light areas when scanning negatives is a lot less noticable. Yes
your right in that the negatives do suffer from noise across wider
density ranges.

If I ever replace the Canon fs2710 then I would look at scanning slides
again and be a little more carefull with exposures.

I believe that FS2710 simply does not have enough dynamic range, or low
enough dark current, to do dark slides. At least in my experience. On
occasion, I intentionally overepxose slides that I intend to scan. It
works, a kind of, and is not that objectionable with today's oversaturated
films.

Pavel

 

[Don]

Are you sure that this is not a problem of the hardware?

Yes, hardware is the cause, but VueScan is the problem. Namely:

Misalignment does happen with all scanners because scanner mechanics
just aren't capable of such high precision i.e. going back to exactly
the same place as during the previous pass. This can be ameliorated
afterwards by sub-pixel alignment in software.

The problem with VueScan (one of many) is that is makes empty claims
without taking this into account. Since no sub-pixel alignment takes
place the end result is a blurry mess. As outlined earlier, you can
get much better results by simply applying a small amount of Gaussian
Blur to dark areas.

To see how different exposure are blended by professional software
check:

http://www.ict.usc.edu/graphics/HDRShop/

Version 1 is free. It has its shortcomings (8-bit input/output,
clipping) but it's definitely worth a try.

My otherwise
miserable CanoScan FS2710 does not show any noticeable bluring with the
long exposure pass

I'm not familiar with that scanner but are you sure you're not
*single-pass* multiscanning?

If it is a genuine *multi-pass* multiscan, then it depends on how you
do the tests. There are mathematical models to do this properly, but
even with a simple visual test you need to follow some basic
guidelines. For example, examine the *two* images at maximum
resolution, overlapped and then toggling between the two. This will
clearly show the amount of misalignment and areas where this is most
apparent. Only then can you go to the combined image and examine
(again at maximum resolution) the areas where the misalignment is most
apparent.

But it could very well be that the blurring doesn't bother you, in
which case you can just ignore all of the above. I mean, if you're
happy with the result then there's no problem. However, anyone who
does care for such detail and level of quality should definitely stay
away from VueScan as far as possible.

Don.