Newbie scanner questions

[hellman]

Not too long ago, I bought a Konica-Minolta DiMAGE 2 scanner to
digitize my slides and print negatives. After a small learning curve,
I'm very happy with the scanner and the results, but have a question
I hope someone here can help with.

With an optical resolution of 5400 dpi and a dynamic range of 4.8, it
should be able to capture essentially all the information on either a
slide or a negative. (Correct me if I'm wrong.) So, when I scan,
should I use automatic exposure and possibly manual over-rides on
exposure or (as it now seems to me) is it better to use fixed exposure
and correct exposure digitally afterward, for example in PhotoShop
Elements?

I'm finding that, even with automatic exposure (which I'm now
doubting I should use), PSE significantly improves many scans. For
example, using Auto Levels often turns a so-so scan into a great one.
Of course, many of PS's tools only work after converting the 16 bit
per channel scan into an 8 bit one, so at that point I've got much
less dynamic range than advertised. (I'm using PSE 3.0 if that
affects things. I could upgrade if it would make a big enough
difference.)

Also, if anyone wants to critique it (again, I'm learning), I've
been scanning at 2700 dpi which produces about 9 megapixel scans. But
each of those is 50 MB in size in TIFF format. I might use the higher
5400 dpi setting on a few, very special photos, but 2700 seems to
provide excellent results, allowing significant cropping or zooming
with great detail. And 5400 dpi would produce 200 MB TIFFs that I
suspect would be slow to open even on my 2 GHz G5 iMac. Along these
lines, I'm really glad I waited til now to start scanning. A few
years ago, even if the scanners were as good (which I doubt), the
storage and computing requirements would have made the process a real
pain. Today, at 9 megapixels per scan, the speed is reasonable.

Thanks for any help.

Martin

 

[Stephen Henning]

So, when I scan,
should I use automatic exposure and possibly manual over-rides on
exposure or (as it now seems to me) is it better to use fixed exposure
and correct exposure digitally afterward, for example in PhotoShop
Elements?

You basically answered your own question. You want to use an exposure
that will keep the most important part of the picture in the center of
the dynamic range. If you want to bring out highlights or shadow areas,
you have to choose exposures that will permit this. You need to have
the most important information where it won't be truncated when your
photo program reduces the dynamic range. I feel a fixed exposure is
completely wrong except for groups of similar slides. Auto exposure is
the best place to start with sometimes manual over-ride used to move the
dynamic range.

 

[hellman]

You basically answered your own question. You want to use an exposure
that will keep the most important part of the picture in the center of
the dynamic range. If you want to bring out highlights or shadow areas,
you have to choose exposures that will permit this. You need to have
the most important information where it won't be truncated when your
photo program reduces the dynamic range. I feel a fixed exposure is
completely wrong except for groups of similar slides. Auto exposure is
the best place to start with sometimes manual over-ride used to move the
dynamic range.

Thanks for the response. In between this and my earlier post, I
discovered that KM's stated Dynamic Range of 4.8 with "theoretical" in
parenthesis is almost surely an overstatement. Apparently, some scanner
makers started giving a "theoretical" DR based merely on the number of
bits output by the scanner, with no regard to the actual DR of the
sensor. Really useless information, and misleading to call it
"theoretical." But according to what I found, once one maker started
doing this, others almost had to follow suit to avoid looking worse.
Does anyone have an idea of the actual DR of my KM DiMAGE 5400-2 and
similar scanners?

 

[Don]

With an optical resolution of 5400 dpi and a dynamic range of 4.8, it
should be able to capture essentially all the information on either a
slide or a negative. (Correct me if I'm wrong.)

That's just theory. In practice all consumer scanners struggle with
slides, most of all with dense Kodachromes. Even my LS-50 with "only"
14-bits (and therefore 1.5-bits overhead) should in theory cover the
full dynamic range of a slide (~12.5 bits) but it does not.

So, when I scan,
should I use automatic exposure and possibly manual over-rides on
exposure or (as it now seems to me) is it better to use fixed exposure
and correct exposure digitally afterward, for example in PhotoShop
Elements?

Oh, you should pull out as much as you can from the film. All editing
afterwards is just "pretend". You can't recreate the missing dynamic
range in software (i.e. by editing) after the scan. So definitely
fine-tune exposure to each image separately.

I'm finding that, even with automatic exposure (which I'm now
doubting I should use), PSE significantly improves many scans. For
example, using Auto Levels often turns a so-so scan into a great one.
Of course, many of PS's tools only work after converting the 16 bit
per channel scan into an 8 bit one, so at that point I've got much
less dynamic range than advertised. (I'm using PSE 3.0 if that
affects things. I could upgrade if it would make a big enough
difference.)

OK, a number of things here...

First of all, it's important to realize what is actually happening
when you scan. In broad strokes, it's a two step process. One, you
obtain the image from film. Two, you edit this image.

Actually, strictly speaking, editing has nothing to do with scanning.
It's something you do *after* the scan. Even when you use various
editing options in scanner software this is no different to scanning
"raw" and then editing that in any image editing software of your
choice afterwards.

Therefore, it's essential to pull as much as possible out of film at
the scanning stage before you even start editing. If you're trying to
get maximum quality (i.e. most data) out of film there are two
concepts you should look into, one is "scanning raw" and the other is
"twin scanning" or "high dynamic range" (HDR) images.

Scanning "raw" means you don't do any editing in your scanning
software. Instead, you only use the settings which affect the
hardware. Essentially this boils down to focus, exposure and ICE (when
applicable). All other settings (contrast, curves, etc) should be
turned off or set to neutral (flat curves for example). Such a scan at
maximum bit depth of your scanner (e.g. 16-bits) is also known as your
"digital negative". Normally you would back this up to DVD and archive
in a safe place. Ideally back it up twice so when one DVD deteriorates
you can "refresh" your backups.

After that you would edit a copy of this image for "consumption". This
is equivalent to making a print in the analog world. This image will
then be targeted to whatever output you're aiming at (screen, print,
etc). This target device will determine final bit depth, resolution
(i.e. image size), file format (e.g. JPG to upload to your web site or
share with friends and family) etc.

At a later date when you get a new bigger monitor or a new printer
with more resolution you go back to your digital negative and create
another copy for this new device. This is not only easier on both film
(no danger of accidentally scratching it during 2nd scan) and the
scanner (no need to scan again) but it's also faster (i.e. you just
copy your digital negative from the DVD). But most of all, analog film
constantly deteriorates. So by scanning "raw" at maximum capabilities
of your scanner and archiving you "freeze" further deterioration.
Unlike analog film data, the digital file data doesn't deteriorate.
The media (DVD) may deteriorate but the data does not!

And so we come to HDR images. As you've discovered scanning slides is
like trying to cover yourself with a short blanket. If you get
highlights properly exposed, the shadows are all noisy. If you get the
shadows properly exposed, the highlights are all blown. The idea of
HDR is to make multiple scans. Usually two scans are enough (one for
shadows and one for highlights) but you can also do multiple scans at
different exposure until you cover the full range.

After that you combine all these images taking the best of each one to
create a composite image where all sections of the dynamic range are
properly exposed. This is very tricky for a number of reasons. Due to
inaccuracies of consumer scanners no two scans will line up perfectly!
Therefore most HDR software comprises *sub-pixel* alignment. Secondly,
"characteristic curves" of film (i.e. scanner/film "behavior" at
different exposures) is not linear! This means that all three channels
do not behave the same way as you change the exposure! Therefore you
need to "color coordinate" (my term) multiple scans before you
combined them! Again, HDR software does this automatically.

Obviously such a composite image has more dynamic range than a "plain
vanilla" 16-bit file format can handle which is why they use 32-bits.
However, you can also reduce such an image to 16-bit for further
processing with conventional software. Again, HDR software does this
automatically.

Of course you should do all your editing in 16-bit after that
(although some HDR programs will edit in 32-bit space as well, as do
newer Photoshop versions) and only covert to 8-bit before the final
step (print or conversion to JPG).

OK, I better stop now... ;o) Check the archive of this group for more
using the key phrases sprinkled above.

Don.

P.S. A couple of interesting free programs you may want to try:

HDRShop:
http://www.debevec.org/HDRShop/
only works with 8-bit input but it's a good start to get your feet
wet. It will read 16-bit files but it converts them to 8-bit before
processing.

Also, download the plugin called "Reinhard HDR Tonemapping Plugin" to
export the combined image!

A free Photoshop plug-in of a 12-bit histogram program:
http://www.reindeergraphics.com/optipix/

Some interesting info about HDR here:

http://www.cgg.cvut.cz/~cadikm/tmo/

etc. Chase up the links in above URLs for more.