Canoscan FS4000, Vuescan, and dust removal

[the_applegates]

Is anyone else using Vuescan with their Canon Canoscan FS4000 film
scanner? If so, how is the dust removal working for you? If you're
using a version in which the dust removal is working well, I would like
to hear about it! And, maybe get a copy of it to try. What I'm
finding in more recent versions is inadequate removal of dust on the
"light" setting, accompanied by a strange accumulation of black
"pepper" (for lack of a better word) surrounding any defects and also
sprinkled lightly all around. (Don't confuse this with so called
peppergrain - its not that, but something completely different). On
the "medium" setting the artifacts go away, but there is a small but
noticeable blurring of the image (unacceptable to me).

I spent a while documenting the artifacts I was seeing and going back
and forth with Ed about it. Unfortunately, in the end I wasn't able to
convince him that there was anything wrong and he basically just
stopped corresponding with me about it. He did confirm that at one
point he adjusted the "strength" of the dust removal at user's request.
But it seems that light is now just too light and produces artifacts,
while medium may soften the image more than I would.

Anyway, to make a long story short, what I have saved right now are the
following versions:

8.2.17 (may be ok)
8.2.34
8.2.36
8.2.37

I only recently realized that I should be saving the Vuescan.exe files
separately! But I'm pretty sure that at one time the dust removal was
working much better, with "light" being just enough without creating
any artifacts. But I just don't have that version anymore!

So, anyone have a "good" version you could send me to try?

Also, I can see if I can upload to somewhere the comparison images I
made to document this issue. Just have to find a place to put them
first!

Thanks,

Jeff

 

[Roger S.]

Hey Jeff,
I was finding that dust removal didn't work well in the post .17
versions with negatives (slides seemed okay).
I did upgrade to 8.2.01 and things seem good- I only scanned a few
negatives, but it seems to be working fine again, with minimal
clean-ups from me.
If I were you, I'd give the new version a try. Also, I never bother
with anything heavier than light cleaning.
Roger

 

[the_applegates]

Roger,

Thanks, and yes I'll have to try the latest version. There's no risk
now that I finally wised up and started saving each version separately!
If you still have the download executable for 8.2.01, could you email
it to me?

I'm with you - I always just used the light setting. This (in the old
days at least) was good enough for me since most of what I was scanning
was fresh film without a lot of dust.

Also, just to be clear, the artifacts I'm talking about actually make
the "light" setting worse than "none". That's my problem.

Jeff

 

[Roger S.]

Hmm, I meant 8.3.01 actually (newish version). I deleted a bunch of
old versions. I do have 8.1.42 but don't remember it very well.
I think version tracker and other websites store archives of Vuescan.

I did some testing of the 8.2.20 series with IR and didn't notice
speckles. I did notice a lot of flaws in the negative not getting
fixed, however!

 

[Steven]

Thanks, and yes I'll have to try the latest version. There's no risk
now that I finally wised up and started saving each version separately!
If you still have the download executable for 8.2.01, could you email
it to me?

I have the 8.2.02 zip file. If you want it, please ask and I can make
it available.

-- Steven

 

[DenverDad]

Well I just tried the latest version, 8.3.03, and I'm seeing the same
exact effects.

As soon as I can find a place to upload them, I'll put up some examples
and studies I did to show the effect. I should add that it is rather
subtle in a way. You have to do some pixel-peeping to see it at first,
maybe at 200% or so. Now I know, that makes it sound like its a pretty
minor issue, but stay with me because I think its not! First of all,
the appearance of dust spots and imperfections actully look WORSE on
the "light" setting than they do with "none". And since the "medium"
setting begins to blur, it basically makes the dust removal almost
unusable. Secondly, I'm concerned that having so many pixels changed
to dark/black (or white in the case of slides) may be having an adverse
effect during post processing - namely with setting white and black
points, anything having to do with clip levels and possibly color
balance. It may also be adding an extra burdon to noise removal
algorithms (e.g., Neatimage).

I'll work on getting some examples up.

Also, Roger that version tracker you suggested looks promising for
getting older versions. It implies I can download any older version
that it shows. But unfortunately when I click on the link, it just
ends up taking me to the current Vuescan page. Am I doing something
wrong (like maybe I need to register first?), or is there really
nothing there?

Jeff

 

[DenverDad]

Ok, here's some links for examples:

The first is a typical example comparing the "none", "light", and
"medium" dust removal settings for the same scan:

http://www.pbase.com/denverdad/image/50656776/original

The next example shows "none" and "light" settings for a crop in which
there is no dust at all. There is nothing to clean, but turning on
light dust cleaning adds the "pepper" noise artifact I've been talking
about.

http://www.pbase.com/denverdad/image/50656638/original

And this one shows how two different versions handle cleaning a couple
of dust blobs from the same scan:

http://www.pbase.com/denverdad/image/50656794/original


Jeff

 

[Roger S.]

Hi Jeff,
Interesting...

I inspected to negative scans I did with 8.3.01 and don't see any
speckles at 300%. I will try to scan something else soon and see if I
have the same problem. Are you scanning slides, negatives, or both,
and do you always have problems? Do you scan at 64 bit, 4000dpi?
Sharpening, etc off?
My 8.3.01 looks like your 8.2.17. Are you sure it's speckled with
8.3.03? Did you try the same negative?

I think you're right about version tracker. There is another website I
read about somewhere here where you can download old versions, but I
forget the name. I ended up doing google searches at one point to find
old versions.

 

[Wilfred]

Also, Roger that version tracker you suggested looks promising for
getting older versions. It implies I can download any older version
that it shows. But unfortunately when I click on the link, it just
ends up taking me to the current Vuescan page. Am I doing something
wrong (like maybe I need to register first?), or is there really
nothing there?

You may also want to try the www archive:
http://web.archive.org/web/*/http://www.hamrick.com
Unfortunately it has no records from December 2004 onwards.

 

[Don]

Well I just tried the latest version, 8.3.03, and I'm seeing the same
exact effects. ....
the appearance of dust spots and imperfections actually look WORSE on
the "light" setting than they do with "none". And since the "medium"
setting begins to blur, it basically makes the dust removal almost
unusable.

Another one for the list! ;o)

Don.

 

[DenverDad]

My scans have been both negatives and slides. The effect seems most
evident with my negative scans where I am seeing black "defects", as
opposed to slides which generate white defects. Almost always at
48bits, 4000dpi. No sharpening (not at scan time anyway - some of my
demo images have been sharpened). And, using a USB connection.

It's interesting you're not seeing the effect, Roger - it should be
pretty evident at 300%. When I first contacted Ed about this my goal
was to first find out whether it was a hardware or software issue. I
thought I had eliminated the hardware possibility. But if you and
other's can't duplicate it I may have to reopen that case.

When I refer to scanning with the latest version, I have to admit to
"cheating" somewhat. Rather than performing a new scan I just brought
up a saved raw file (i.e., 64bit RGBI) - the same one I sent to Ed to
have a look at in fact - and opened it up in Vuescan (nice to have this
feature by the way). My thinking was that the cleaning algorithm must
be applied to the raw RGB scan using the I (infrared) data, so as long
as I had raw RGBI scan data, it wouldn't matter if I used an old raw
file or if I created one with a new scan. But who knows? Maybe it
does! I will go back and do a proper scan to be sure about whether the
latest version really shows the problem or not.

Jeff

 

[Roger S.]

"My thinking was that the cleaning algorithm must be applied to the raw
RGB scan using the I (infrared) data, so as long
as I had raw RGBI scan data, it wouldn't matter if I used an old raw
file or if I created one with a new scan. "

Hmm, it shouldn't make a difference, but I think VS was having some
alignment issues with the IR channel in past versions. It's worth
trying on a fresh scan. I'll also poke around when I get a chance.
I've only scanned a few strips of negatives with 8.3.01, but haven't
noticed anything, and I do examine the scans closely for flaws.

Do you calibrate before you scan? I set the calibration period to 1-
so it calibrates each day.

 

[Steven]

Another one for the list! ;o)

Don.

Wow, now we have at least two 'bugs' in VS. Heaven help Microsoft if
Donny ever starts on XP.

Donny is bugged and unreasonable - check the archives for details.

-- Steven

 

[DenverDad]

Just scanned the same image in with the latest version (8.3.03). Same
basic results. Check out the latest:

http://www.pbase.com/denverdad/image/50691836/original

There's basically two goobers in this little crop. It's interesting
that the one which looks like a fiber seems to attract the "pepper",
but the other one (the big "blob") doesn't - it looks like its
corrected more normally.

Yes, I calibrate most every time also.

Jeff

 

[Roger S.]

Well, thanks for the clarifying post. The blobs are as craptacular as
I've come to expect from VS cleaning, but the little black dots all
over the place are inexplicable.

I just made a 4000dpi scan of a neg (lovely Max400) and compared light
to none. The cleaning worked as expected and there are no dots that I
can see comparing the scans side by side at 100%. I enlarged around
the problem areas and cannot see any evidence of the problem you are
faced with.

Can you reset the ini to default and give it another whirl? Keep the
scanner profile to built-in.

Also, VS 8.3.04 is out now with improved infrared cleaning. I haven't
tried it.

 

[DenverDad]

"Also, VS 8.3.04 is out now with improved infrared cleaning "

Maybe Ed has been listening!

Sure, I'll try the default ini, then I'll try the new version.

May take a bit!

Jeff

 

[DenverDad]

Update: I've spent the last couple days playing with VueScan version
8.3.04 and I can report that I'm not seeing any of the little pepper
artifact I was seeing before. So basically dust removal is working
again! Well, at least as well as before, anyway. The "light" setting
still seems to be a bit underpowered sometimes, but at least there's
none of the pepper.

I still have no explanation for why others weren't seeing this. Maybe
it has something to do with registration between the RGB and the IR
scans. In any event, I've got a usable system now, so I'm happy.

Jeff

 

[Roger S.]

Glad to hear it, Jeff! I'm still on 8.3.01 and it works well enough.
I've been scanning silver B&W negatives recently, so I haven't used the
IR cleaning. Having a scanhancer is quite handy for reducing visible
scratches and dust.

 

[DenverDad]

Talk about putting two and two together... I just realized that you
must be the same Roger I talked to awhile back regarding the
scanhancer. You sent me some sample images (with and without the
scanhancer). I almost ordered the thing, but In the end I decided not
to, figuring that the very slight improvement wasn't worth it for me in
terms of the extra scan time it created. But I guess that's a topic
for another thread!

Jeff

 

[Roger S.]

Yes, and you must be the Jeff Applegate from Photo.net I keep running
into : )

The scanhancer is a time-saver for B&W and not really worth it for
negative or chromes in my opinion. It does keep Vuescan from
overcleaning slides and eating away at high-contrast edges, but this is
almost unnoticable.

For B&W you only do one pass, and while this may be twice the length,
you save time on retouching because there's a lot less dust and
scratches.