Viewscan: now this is a weird one...

[Don]

I'm sure you're not being even a tad patronising with the tone of the above
comments, so I'll regard them as constructive - on which basis, thanks for
your input.

Indeed! That's why I stressed that I can't speak for your requirements
and only focused on facts.

Objective facts can never be patronizing. Whenever these facts are
subjectively perceived as such, it's in the eye of the beholder, as
the saying goes.

As I'm sure you stressing "constructive" in your original question was
not a jab aimed at anyone in particular but a genuine request for
meaningful and serious feedback which is why I responded.

But no, my needs are in fact quite demanding (I'm a pro photographer), and
as I can't upgrade my hardware just yet, I do need to get the max from what
I have. I'm assuming (maybe wrongly) that the optics were the main limiting
factor with the scanner. But maybe not.

Optics certainly play a part but that's beyond our control. The next
step is firmware, and that's also beyond our control. On top of all
that is the scanner software and there we do have a choice.

As I wrote elsewhere (and please ignore if you're already aware of
this, as I suspect you are) the common misperception is to judge the
scanner software by its editing features i.e. how the scan "looks".
The point is, scanner software is composed of two *unrelated*
components: the actual scanning (i.e. getting data from the scanner)
and editing tools.

These editing tools are a poor substitute (i.e. a very limited
quick-and-dirty subset) of standalone external editors like Photoshop.
They exist only for marketing reasons and amateurs for whom that's
good enough and they can't justify the cost of a dedicated editor or
wish to spend the extra time.

Since (I think) that's not the case here, the acid test then is how
good is scanner software at its real task i.e. getting the data from
the scanner. And there VueScan fails miserably. It corrupts data left
and right. Furthermore, its notorious track record regarding bugs
means it just can't be trusted. Also, each new release shuffles the
deck completely as new bugs are introduced and old ones are
reactivated. All that makes VueScan totally unsuitable for anything
other than casual use for a small print or a tiny Web JPG.

This "moving target" aspect makes it also very difficult to do
comprehensive testing because the whole testing suite must be repeated
for each new version. When I tested VueScan a couple of years back I
was appalled at the results and dismissed it. Granted, I have not
repeated them since but given its track record and endless complaints
reason dictates nothing much has changed. If anything, it's worse.

So to get back to your question, I suspect (and correct me if I'm
wrong) as a pro you scan "raw" and do all your editing in
post-processing (i.e. in Photoshop). Therefore, the only thing you are
concerned with is getting the purest data possible from the scanner,
uncontaminated by anything. Presumably, the intention is to archive
this as your "digital negative" and then work on a copy for
"consumption" (printing, Web, etc). The usual pro/semipro workflow.

If that's the case, then you do need to test VueScan at that most
fundamental level (raw). Instead, you seem to make an assumption "I
always scan to a (theoretically) lossless .TIFF". Now, that's a
dangerous assumption to make at the best of times, but especially so
in case of VueScan. Yes, the *format* (TIF) itself is lossless but
that's beside the point here because it's about *the data* you put
into that format that should be of main concern!

And that was the main gist of what I wrote last time (see quote on
top). Maybe it didn't come across well but should be clear now.

And, having paid for ViewScan Pro, plus Photoshop, I'd kind of hoped I was
reasonably well-equipped. Certainly, the results are WAY better than with
the supplied Canon software. I can live with the ViewScan interface; works
for me, now I'm used to working with it, with plenty of (advanced options)
tweaks available.

Again, it's not entirely clear what your workflow is, but the above
appears to suggest you *are* editing in your scanner software after
all. If that's the case, then you certainly will not be able to see
the corruption at the most fundamental level because it's masked.
(That's why I wrote "depends on how closely you look".)

From that I can only conclude (without being patronizing but just
drawing a logical conclusion) that you are not evaluating the *data
acquisition* part of your scanner software but only looking at the
final product i.e. the image and making a visual (subjective)
assessment.

Of course, there's nothing wrong with that per se, but that doesn't
say anything about how good the software is at data acquisition. And
no matter how appalling Canon software may be in other areas (i.e.
editing) and even without knowing it first hand, common sense is that
it's better than VueScan. Why? For at least two reasons: Canon
certainly knows their hardware better than any 3rd party and VueScan's
notorious track record. Even if we question the first reason, there's
no doubt about the second. And that alone is enough.

That said, what would you suggest for a better-performing package, and why?

As often mentioned elsewhere the only other option appears to be
SilverFast (http://www.silverfast.com/). It's pricey but considering
you're in the business and can afford Photoshop you should be able to
afford it as well (it's tax deductible, anyway).

Finally, given the caveat that I don't have any personal experience
with Canon scanners and assuming the above workflow (raw scan =>
digital negative => archive => edit) I'm fairly confident that in
spite of frequent complaints it will be able to get a decent raw scan
out. And it will (consistently!) do so far better than VueScan.

Don.

 

[Kurt Stege]

Moin, moin.

As a forward, I am not really loving Vuescan, but I am using it
because it fulfills my requirements and I don't see any reasonable
alternative.

But the main question is what you expect scanner software to do?

It's very important to realize that there are two distinct components
of scanner software: scanning (i.e. digitizing film data) and
processing (i.e. image editing).

This is a very good point.

Most people confuse this built-in image editing with scanning. This
image editing has absolutely nothing to do with scanning and is there
only as a matter of convenience for people who don't have a separate
editing program. However, any such built-in editing is very limited
because the main purpose of scanning software is to scan, not edit.

Correct. At least, it meats my personal opinion.

And this is, what I expect from part one of my scanning software:
Get an image from the scanner without any further processing,
just as it leaves my scanner hardware.

Question 1: What are the problems of Vuescan for just scanning?
I don't see any real issues. Scanning and saving the result as
"raw file" (in the language of Vuescan) is done well by that software.
The result is a Tiff file with four channels, R, G, B, and IR,
and 16 bit per channel.

Question 2: What other software provides this feature? Please
note, that running some ICE algorithms like evaluating the IR
channel for dust hiding is clearly not part of scanning but
part of processing.

And now my trouble begins. For processing the images, I am using
Vuescan as well, but I am not really happy with it. There is so
much incredible Voodoo Magic that changes in uncontrolled way
from version to version. This is the part where the bugs are
coming and going. This is the part where I stuck to a given
fixed version of Vuescan as long as possible. Every few years
I am upgrading to some current version of Vuescan and am trying
a lot of versions until I find one that has the new feature that
I want to upgrade for and performs reasonable well in all other
points.

Question 3: What other software is able to process the raw
image files that results from task 1, from scanning? This includes
processing the IR data for dust and scratch handling. As far as
I know, no OEM software from the scanner manufacturer is able
to process files instead of actual scanner data. Siverfast might
be an answer. Is Photoshop able to do dust removement or does
Photoshop know about typical film behaviour?

Question 4: Assuming that Silverfast is able to do processing
of image files (and not only processing of electrical scanner
data as the OEM software): Can I use Vuescan as a universal
driver for all scanner hardware and one single Silverfast
licence (that is bound to a specific scanner model) to process
that data?

Which then goes back to the above question. If you are after high
quality then you should not be relying on scanner software for
editing. You should really scan raw and do the editing in dedicated
image editing software which has the full complement of tools not the
cut-down and limited sub-set found in most scanner software.

See question 3.

And for that purpose (scanning without editing) virtually all software
supplied with the scanner will do the job very nicely (if you turn
everything else off!). After all if the manufacturer doesn't know the
hardware who does? Where this software usually comes short is editing,
but as I mention above, that's not really its primary job.

But exactly this part, scanning, is something that can be done
stable and convenient by Vuescan as well. Nearly all known (and
unknown) bugs of Vuescan are related to processing, not to scanning.

Whatever the case, VueScan is not really suitable for any serious
scanning. It may be OK for a quick and dirty (tiny) JPG to be posted
on a Web site, but not for anything else.

In this point, I disagree. In my opinion, Vuescan is really suitable
for scanning and under special circumstances suitable for processing.
When you are talking about preparing an image as JPG for a web site,
your are talking about processing, not about scanning.

However, for some people who don't care for quality VueScan may do the
trick assuming they can get to grips with the insane "user interface".
Nevertheless, for such people with a very low threshold I would only
advise to find a version with the fewest bugs (for their use) and
stick with it!
*AOL*

Don't upgrade automatically but (if you must upgrade)
wait for others to identify the bugs first. And - most of all (!) -
*keep* your old version!

Even more, I regularly download the current version, just for my
archive. When I have to upgrade any time for any reason, I have a
great fundus of versions to choose from. I am sorry, but that is
necessary.

Best regards,
Kurt.

 

[Randy Howard]

Kurt Stege wrote

As a forward, I am not really loving Vuescan, but I am using it
because it fulfills my requirements and I don't see any reasonable
alternative.

I'm in the same boat.

And this is, what I expect from part one of my scanning software:
Get an image from the scanner without any further processing,
just as it leaves my scanner hardware.

Question 1: What are the problems of Vuescan for just scanning?
I don't see any real issues. Scanning and saving the result as
"raw file" (in the language of Vuescan) is done well by that software.
The result is a Tiff file with four channels, R, G, B, and IR,
and 16 bit per channel.

More importantly, when you do the saves, the files actually show
up in the correct directory, in the correct file format, and
they are not zero bytes in length. That is something that
Photostudio can't claim with any degree of credibility.

And now my trouble begins. For processing the images, I am using
Vuescan as well, but I am not really happy with it. There is so
much incredible Voodoo Magic that changes in uncontrolled way
from version to version. This is the part where the bugs are
coming and going. This is the part where I stuck to a given
fixed version of Vuescan as long as possible. Every few years
I am upgrading to some current version of Vuescan and am trying
a lot of versions until I find one that has the new feature that
I want to upgrade for and performs reasonable well in all other
points.

Ugh. I don't have a lot of happiness in the future. I'm
starting to think Don is on the right track about chasing down a
bad alley. Anyone want to buy an almost new, pristine Canon
9950F?

Is Photoshop able to do dust removement or does
Photoshop know about typical film behaviour?

I don't know of a plugin that does true ICE-like processing, but
there is that Polaroid "Dust and Scratch" removal software that
was recently mentioned here. It's hyper agressive though, you
really have to tweak the settings to get a usable result.

But exactly this part, scanning, is something that can be done
stable and convenient by Vuescan as well. Nearly all known (and
unknown) bugs of Vuescan are related to processing, not to scanning.

I hope that is the case, it would make it more bearable.

In this point, I disagree. In my opinion, Vuescan is really suitable
for scanning and under special circumstances suitable for processing.
When you are talking about preparing an image as JPG for a web site,
your are talking about processing, not about scanning.

If I can get a suitable tiff file into photoshop, I'll be happy
with it.

Even more, I regularly download the current version, just for my
archive. When I have to upgrade any time for any reason, I have a
great fundus of versions to choose from. I am sorry, but that is
necessary.

Good to keep in mind.

 

[Maris V. Lidaka Sr.]

Since (I think) that's not the case here, the acid test then is how
good is scanner software at its real task i.e. getting the data from
the scanner. And there VueScan fails miserably. It corrupts data left
and right. Furthermore, its notorious track record regarding bugs
means it just can't be trusted. Also, each new release shuffles the
deck completely as new bugs are introduced and old ones are
reactivated. All that makes VueScan totally unsuitable for anything
other than casual use for a small print or a tiny Web JPG.

I don't know what you base this opinion on - Vuescan does not corrupt data
"left and right", and it can be trusted. I and many others do not have the
problems you suggest.

This "moving target" aspect makes it also very difficult to do
comprehensive testing because the whole testing suite must be repeated
for each new version. When I tested VueScan a couple of years back I
was appalled at the results and dismissed it. Granted, I have not
repeated them since but given its track record and endless complaints
reason dictates nothing much has changed. If anything, it's worse.

Not having tested Vuescan for a "couple of years", you should at this point
withhold your opinion. Or, if you want to base your opinion on newsgrop
posts, you must consider both the positive posts and the complaints.
You are well-equipped.

Again, it's not entirely clear what your workflow is, but the above
appears to suggest you *are* editing in your scanner software after
all. If that's the case, then you certainly will not be able to see
the corruption at the most fundamental level because it's masked.
(That's why I wrote "depends on how closely you look".)

Not all "tweaks" constitute editing in the scanner software - they often
constitute adjustments necessary and/or beneficial to obtaining all of the
data available in the film or slide.

As often mentioned elsewhere the only other option appears to be
SilverFast (http://www.silverfast.com/). It's pricey but considering
you're in the business and can afford Photoshop you should be able to
afford it as well (it's tax deductible, anyway).

SilverFast is good software, no doubt better than Canon's.

Finally, given the caveat that I don't have any personal experience
with Canon scanners and assuming the above workflow (raw scan =>
digital negative => archive => edit) I'm fairly confident that in
spite of frequent complaints it will be able to get a decent raw scan
out. And it will (consistently!) do so far better than VueScan.

This I disagree with - Vuescan will do better than Canon, as will
SilverFast.

Maris V. Lidaka Sr.
Riga Company

 

[Don]

Moin, moin.

Jawohl! ;o)

As a forward, I am not really loving Vuescan, but I am using it
because it fulfills my requirements and I don't see any reasonable
alternative.

Fair enough. As I always say: Great! Enjoy!

Correct. At least, it meats my personal opinion.

And this is, what I expect from part one of my scanning software:
Get an image from the scanner without any further processing,
just as it leaves my scanner hardware.

That makes you a member of the so-called "raw scan sect"! I too am a
proud card-carrying member of this unofficial organization! ;o)

Question 1: What are the problems of Vuescan for just scanning?
I don't see any real issues. Scanning and saving the result as
"raw file" (in the language of Vuescan) is done well by that software.
The result is a Tiff file with four channels, R, G, B, and IR,
and 16 bit per channel.

The problems are many. For starters this so-called raw file is not -
as the program claims - raw data straight off the scanner.

This is exactly what I was after and I ran all sorts of tests. That's
not an easy thing to do with VueScan because it does a lot of "secret"
data massaging. Some of it documented, but very hard to locate or
figure out because of the totally weird and messed-up "user
interface". But most of it are just bugs.

For various reasons (mostly trying to hide the bugs) VueScan messes
with this so-called "raw" data. Depending on the scanner and VueScan
version, this may be more or less evident but the rule of the thumb is
this data is bound to be corrupted one way or another.

Of course, that's only the start of all the troubles! If you use this
raw file in the *misleadingly* named "scan from raw" you're opening
yourself to all sorts of other problems. There are "issues" with
profiles being *secretly* applied to this alleged raw file, etc.

Question 2: What other software provides this feature?

Every and any! All you need to do is disable all the editing features
and you will get a raw scan. Some manufacturer's software may be more
difficult to convince to do this but they are all capable of it. For
example, NikonScan's "Auto Exposure" is very "sticky" and quite
difficult to get rid off (you need to close NikonScan *and* turn off
the scanner!).

Technically, for a raw scan you should also scan at linear gamma (1.0)
but most people don't because - unless you edit in linear gamma space,
and virtually nobody does - there's no point scanning in gamma 1.0 if
the first thing you do afterwards is apply gamma 2.2 (of 1.8 for
Macs).

Please
note, that running some ICE algorithms like evaluating the IR
channel for dust hiding is clearly not part of scanning but
part of processing.

That's not entirely true. ICE falls in between. It's part hardware,
and part software.

The trouble is that (for marketing "reasons") you are not able to use
the separate IR channel in ICE even if you saved this IR channel
separately. So, for practical reasons, the only time you can use ICE
is at scan time because of the way ICE is marketed/implemented.

Technically, of course, there's nothing preventing them from letting
you apply ICE after the scan but, as I say, for marketing "reasons"
they decided against it.

And now my trouble begins. For processing the images, I am using
Vuescan as well, but I am not really happy with it.

I'm not surprised.

There is so
much incredible Voodoo Magic that changes in uncontrolled way
from version to version. This is the part where the bugs are
coming and going. This is the part where I stuck to a given
fixed version of Vuescan as long as possible.

As I wrote before, once you find a version that (sort of) does what
you want, you should always resist the urge to upgrade. And if you do,
make sure you keep the old version in case the new one is even worse.

Question 3:

Actually that's many questions in one!

What other software is able to process the raw
image files that results from task 1, from scanning?

Any image editing software can process the raw scan. A raw scan is
just a picture like any other picture. Image editing software doesn't
know (or care) where and how this image has been created.

This includes
processing the IR data for dust and scratch handling.

Any image editing software can do this as well but it just requires
more operator skill. There are many pages on the Web with instructions
how to do this but it's very unlikely you'll get better results that
ICE. At least not automatic cleaning. You can do better if you correct
each fault manually, bur for that you don't need IR data.

Anyway, there used to be a guy around here called Wilfred van der
Vegte who created Photoshop actions to automate this. On his web page
(I can't find right now...) he used to offer to send these actions if
you email him. The last two emails I have are:

Wilfred
e-mail: (first five letters of my name) at gmx dot net

and

<[email protected]>
Replace 'invalid' with my first name to reply by e-mail

I haven't seen Wilfred around here lately but he may be lurking so if
you get no response at the above two emails, write a message here
asking him to comment if he's still reading.

As far as
I know, no OEM software from the scanner manufacturer is able
to process files instead of actual scanner data. Siverfast might
be an answer.

Some OEM software can. In case of Nikon, for example, that's exactly
what the NEF files are for. So if you save your Nikon scan as *.NEF
you can import that back into NikonScan for additional editing,

--- from NikonScan docs ---
NEF (Nikon Electronic image Format)

The NEF file type is the Nikon Electronic image Format file. NEF files
are saved from Nikon Scan software with curve, histogram and other
modifications saved as editable components; the original scan data is
not modified. If you wish to revise a file, changing any of its
characteristics in Nikon Scan, the NEF file can be reopened and the
file modified.
--- end quote ---

As a matter of fact, NikonScan can open any image file it supports and
you can do all the editing using your saved curves, etc. NikonScan can
also read saved Photoshop Curves, etc.

Is Photoshop able to do dust removement

Depends on the person running Photoshop. The program itself doesn't
have any automatic menu options as far as I know - but I use an old
version (6). Maybe new versions do have this as an option.

Also, see above actions by Wilfred.

or does
Photoshop know about typical film behaviour?

I'm not quite sure what you mean by this. Are you referring to the
characteristic film curves? If so, then not really, but neither would
I expect Photoshop to do it. You can get some correction if you use
relevant film profiles.

Question 4: Assuming that Silverfast is able to do processing
of image files (and not only processing of electrical scanner
data as the OEM software): Can I use Vuescan as a universal
driver for all scanner hardware and one single Silverfast
licence (that is bound to a specific scanner model) to process
that data?

As I say, I haven't looked into Silverfast so I'll have to let someone
else reply to this.

If Silverfast can read TIF files then obviously you don't need a
license because Silverfast has no way of knowing where those files are
coming from. In theory they can store this in the TIF header, but
that's easy to "fix".

Either way, if you are after such high quality I would strongly
suggest dropping VueScan. It's really not suitable for that.

But exactly this part, scanning, is something that can be done
stable and convenient by Vuescan as well. Nearly all known (and
unknown) bugs of Vuescan are related to processing, not to scanning.

That's not true! VueScan fails at getting this raw data. That's the
first problem. It just messes up this data. Either directly, in order
to hide other problems. Or indirectly because of a confusing user
interface so things may be done without you even knowing. And then on
top of that there are the many bugs...

In this point, I disagree. In my opinion, Vuescan is really suitable
for scanning and under special circumstances suitable for processing.
When you are talking about preparing an image as JPG for a web site,
your are talking about processing, not about scanning.

You misunderstood. My point was that if the quality requirements are
low: small image (low resolution) and JPG (lossless compression, 8-bit
color). In such an image even VueScan corruption will be hard to see.

But if you do anything else, for example, use native scanner
resolution and bit depth (e.g. 16-bit color) and save in a lossless
TIF format, all those problem become obvious if you examine the image
carefully.

Best regards,

Tschüss! ;-)

Don.

P.S. I used to live in Germany for a while.

 

[Don]

I don't know what you base this opinion on - Vuescan does not corrupt data
"left and right", and it can be trusted.

Your fundamental mistake is it's not an opinion but a substantiated
and objective *fact*! So all your subsequent conclusions based on the
above erroneous assumption are also false.

Add to that the constant stream of complaints by many other users with
specific and *documented* cases of VueScan data corruption. Some of
them are in the partial (!) list I posted the other day, so refer to
that for more details, or just browse the archives.

Finally, considering how unreliable VueScan is with endless bugs all
over the place it's just common sense that the data it produces is
heavily corrupt. This corruption may have different causes in
different versions but there's no escaping the "constant" that the
corruption is virtually always there.

You just can't have so many bugs as VueScan does without them
negatively affecting data integrity!

Now, depending on the VueScan "bug of the day" and your workflow (i.e.
requirements) you may not be bitten by any particular bug (or, to be
exact, notice it!) but on balance, if you use VueScan you are bound to
suffer data corruption. Whether you see it or care, is another matter.

I and many others do not have the
problems you suggest.

Yes, you do! You just don't care so you don't see it! You haven't
examined the data carefully but look at the scan and go: "Good enough
for me! I see no problem".

How do I know this? Because if you did examine the data *objectively*
(and not through rose-colored glasses of your, self-evidently, limited
requirements) you could not make statements like the above.

And then, that's not *objective* testing! That's *subjective opinion*.

Nothing wrong with that, of course. To each his own, as the saying
goes, but...

The problem is when you try to elevate this *unscientific* subjective
feeling into objective fact as you do when you make statements like
the above.

If you disagree, then: What are your specific requirements? What is
your specific workflow? What have to done to confirm that what you
*think* you are getting is what you really are getting?

Not having tested Vuescan for a "couple of years", you should at this point
withhold your opinion.

Not if there are virtually daily confirmations. If anything, VueScan
keeps getting worse. Again, see the list!

Or, if you want to base your opinion on newsgrop
posts, you must consider both the positive posts and the complaints.

Absolutely! However, there are no positive posts with objective facts.

The problem is all of the "positive" posts are of the type: "I like
it" or "I see no problem". etc. No data. No specifics. Just personal
*feelings* without any objective frame of reference whatsoever!

All the negative posts are very specific i.e. objective *facts*!
Specific bugs. Specific errors. Specific consequences. Again, just see
the list!

Not all "tweaks" constitute editing in the scanner software - they often
constitute adjustments necessary and/or beneficial to obtaining all of the
data available in the film or slide.

Clearly that's not what I'm referring to in the above paragraph. So
you're missing the point.

But, never mind. I'll bend over backwards and take those non-editing
(!) adjustments e.g. exposure. VueScan has had an exposure bug whereby
it levels off at a certain point and incorrect exposure is reported
back. Essentially, the actual exposure used is unknown and arbitrary.
Furthermore, this exposure corruption is *silent*!!! You don't even
know it's happening unless you do proper testing!

So, any way you cut it VueScan is just a case of amateur bungling.

But - I hasten to add - be that as it may, it can still satisfy some
user's needs (if they have low requirements). And - again - as I
always say: More power to them! As long as they don't try to turn
those low-level subjective feelings into absolute objective fact.

SilverFast is good software, no doubt better than Canon's.

I have no opinion on SilverFast. It's exactly the opposite
(conceptually) of what I'm after so I didn't spend too much time
testing it. All I can say is it appears to do what it's supposed to. I
haven't noticed any blatant problems. I don't like their marketing
strategy (scanner-specific versions) but that's another thing...

This I disagree with - Vuescan will do better than Canon, as will
SilverFast.

As I said last time, I don't know because I don't have the scanner.
But common sense dictates that if you turn all the "extras" off I just
can't conceive that the manufacturer software would fail to deliver.
Anything's possible, of course, but it seems highly improbable.

By contrast, it's a well established fact that VueScan heavily
corrupts the data, no matter what.

Don.

 

[Roger S.]

As someone with a Canon scanner, the default drivers do fail to
deliver, especially for chromes. You can't get a "raw scan" or a
linear gamma scan, so you can't use a scanner profile with raw data,
the IR algorithm gives bizarre results with certain negatives, etc.

You may be better off with Vuescan *if* the version you are using
doesn't have any bugs that unacceptably affect image quality (I can put
up with UI bugs and annoying workarounds). Determining whether a given
version works or not takes testing and a reference for comparison
(stock driver/Silverfast/old Vuescan version). Trust but verify.

"By contrast, it's a well established fact that VueScan heavily
corrupts the data, no matter what."
Right. Image editing also "heavily corrupts" raw data, but so long as
you are in control of it, who cares where you start the editing
process?

The 16 bit 4000dpi tiffs I get from Vuescan with custom color balance
done in-program are a good starting point for my editing and have no
visible posterization, banding, highlight clipping, or other artifacts.
There is no evidence to support the statement that corruption is
"constant." It stands to reason if some versions are defective, but the
bugs are continually fixed, then there must be some working versions
before new bugs are (re)introduced.

I only bother with raw scans on long nightime exposures, where it's not
helpful to do any semi-automated color adjustments in the scan program,
and I need every last bit of unmutiliated data (efficient workflow be
damned.) Your mileage will vary. Have fun.

 

[Roger Moss]

Indeed! That's why I stressed that I can't speak for your requirements
and only focused on facts.

Objective facts can never be patronizing. Whenever these facts are
subjectively perceived as such, it's in the eye of the beholder, as
the saying goes.

As I'm sure you stressing "constructive" in your original question was
not a jab aimed at anyone in particular but a genuine request for
meaningful and serious feedback which is why I responded.


Optics certainly play a part but that's beyond our control. The next
step is firmware, and that's also beyond our control. On top of all
that is the scanner software and there we do have a choice.

As I wrote elsewhere (and please ignore if you're already aware of
this, as I suspect you are) the common misperception is to judge the
scanner software by its editing features i.e. how the scan "looks".
The point is, scanner software is composed of two *unrelated*
components: the actual scanning (i.e. getting data from the scanner)
and editing tools.

These editing tools are a poor substitute (i.e. a very limited
quick-and-dirty subset) of standalone external editors like Photoshop.
They exist only for marketing reasons and amateurs for whom that's
good enough and they can't justify the cost of a dedicated editor or
wish to spend the extra time.

Since (I think) that's not the case here, the acid test then is how
good is scanner software at its real task i.e. getting the data from
the scanner. And there VueScan fails miserably. It corrupts data left
and right. Furthermore, its notorious track record regarding bugs
means it just can't be trusted. Also, each new release shuffles the
deck completely as new bugs are introduced and old ones are
reactivated. All that makes VueScan totally unsuitable for anything
other than casual use for a small print or a tiny Web JPG.

This "moving target" aspect makes it also very difficult to do
comprehensive testing because the whole testing suite must be repeated
for each new version. When I tested VueScan a couple of years back I
was appalled at the results and dismissed it. Granted, I have not
repeated them since but given its track record and endless complaints
reason dictates nothing much has changed. If anything, it's worse.

So to get back to your question, I suspect (and correct me if I'm
wrong) as a pro you scan "raw" and do all your editing in
post-processing (i.e. in Photoshop). Therefore, the only thing you are
concerned with is getting the purest data possible from the scanner,
uncontaminated by anything. Presumably, the intention is to archive
this as your "digital negative" and then work on a copy for
"consumption" (printing, Web, etc). The usual pro/semipro workflow.

If that's the case, then you do need to test VueScan at that most
fundamental level (raw). Instead, you seem to make an assumption "I
always scan to a (theoretically) lossless .TIFF". Now, that's a
dangerous assumption to make at the best of times, but especially so
in case of VueScan. Yes, the *format* (TIF) itself is lossless but
that's beside the point here because it's about *the data* you put
into that format that should be of main concern!

And that was the main gist of what I wrote last time (see quote on
top). Maybe it didn't come across well but should be clear now.


Again, it's not entirely clear what your workflow is, but the above
appears to suggest you *are* editing in your scanner software after
all. If that's the case, then you certainly will not be able to see
the corruption at the most fundamental level because it's masked.
(That's why I wrote "depends on how closely you look".)

From that I can only conclude (without being patronizing but just
drawing a logical conclusion) that you are not evaluating the *data
acquisition* part of your scanner software but only looking at the
final product i.e. the image and making a visual (subjective)
assessment.

Of course, there's nothing wrong with that per se, but that doesn't
say anything about how good the software is at data acquisition. And
no matter how appalling Canon software may be in other areas (i.e.
editing) and even without knowing it first hand, common sense is that
it's better than VueScan. Why? For at least two reasons: Canon
certainly knows their hardware better than any 3rd party and VueScan's
notorious track record. Even if we question the first reason, there's
no doubt about the second. And that alone is enough.


As often mentioned elsewhere the only other option appears to be
SilverFast (http://www.silverfast.com/). It's pricey but considering
you're in the business and can afford Photoshop you should be able to
afford it as well (it's tax deductible, anyway).

Finally, given the caveat that I don't have any personal experience
with Canon scanners and assuming the above workflow (raw scan =>
digital negative => archive => edit) I'm fairly confident that in
spite of frequent complaints it will be able to get a decent raw scan
out. And it will (consistently!) do so far better than VueScan.

Don.

Sorry to breach etiquette and repost the above in its entirety, but it shows
that there is still some intelligent life out there among the morons with
more time than brains.

Point(s) taken, and I'll pursue the RAW option to see what is coming out
without any decision-making by the scanning software. I always shoot my
digital in RAW, and don't need any further convincing that this is indeed
the way to go.

Thanks again.

RM

 

[Don]

As someone with a Canon scanner, the default drivers do fail to
deliver, especially for chromes. You can't get a "raw scan" or a
linear gamma scan, so you can't use a scanner profile with raw data,
the IR algorithm gives bizarre results with certain negatives, etc.

A number of things here...

As mentioned before, virtually all people who scan raw don't use
linear gamma because very few have their system set up in linear
gamma. So the first thing they must do before editing is to convert to
their monitor gamma. Therefore, there is very little point scanning in
linear gamma. So, for practical reasons, the inability of Canon
software to scan in linear gamma is really only of academic interest
and has not practical (negative) effects.

Scanner profiles. I don't quite get what you mean you "can't use
them"? I don't follow the logic. Can you be more specific? For what
it's worth, scanner profiles have very limited usefulness (unlike
monitor and printer profiles). As I've written in detail in the past,
more often than not scanner profiles can actually cause damage (e.g.
if the final editing goes against the profile).

I'm also not sure what IR algorithm you're referring to. If you're
talking about ICE then you can't use it with silver based negatives
(B&W or Kodachromes). If you're talking about VueScan's so-called "IR
cleaning" then all bets are off because of the long standing VueScan
IR bugs.

To wrap up, if you turn off all editing options in your Canon software
(whatever they are) or, failing that, leave all settings "in neutral"
e.g. flat curves, then what you get is effectively a raw scan. True,
Canon software may still be "doing stuff" but this will be kept to a
minimum. It will also be consistent (unlike VueScan's "bug du jour"
corruption) and therefore easier to edit in postprocessing.

You may be better off with Vuescan *if* the version you are using
doesn't have any bugs that unacceptably affect image quality (I can put
up with UI bugs and annoying workarounds).

The problem is VueScan corrupts data at the lowest possible level so
it affects *all* images. The amount of this corruption will vary
depending on VueScan version and the scanner in question.

Do note that this says nothing about how acceptable or unacceptable
this level of corruption may be (that's a subjective call each user
must make). It merely states the fact that all data out of VueScan is
corrupt by definition.

As to UI bugs, this is not as harmless as you make it out to be. Yes,
you may learn to put up with them but not the damage they are hiding.
For example, the exposure bug whereby the given exposure is rolled
back (silently!) and the actual exposure is unknown. That's not just
annoying but damaging. The fact that this occur silently is doubly
damaging! And there are plenty more of that in VueScan.

Determining whether a given
version works or not takes testing and a reference for comparison
(stock driver/Silverfast/old Vuescan version). Trust but verify.

Yes, some conclusions may be drawn from a head-to-head comparison
(given enough detail) but that's second hand at best. This is because
the programs you are comparing to may have problems themselves (e.g.
comparing to older VueScan versions is totally pointless). Therefore,
the only comparison that is reliable and meaningful is a direct
comparison against a known reference.

"By contrast, it's a well established fact that VueScan heavily
corrupts the data, no matter what."
Right. Image editing also "heavily corrupts" raw data, but so long as
you are in control of it, who cares where you start the editing
process?

You're missing the point. With editing you have a choice. You chose
how much editing (i.e. corruption) to apply. With VueScan you don't
have a choice because its corruption is outside of your control. It's
fundamental i.e. below the settings. That's the problem!

The 16 bit 4000dpi tiffs I get from Vuescan with custom color balance
done in-program are a good starting point for my editing and have no
visible posterization, banding, highlight clipping, or other artifacts.

In other words, you've stated your personal preference.

I say nothing about anyone's personal (subjective) preference.

There is no evidence to support the statement that corruption is
"constant." It stands to reason if some versions are defective, but the
bugs are continually fixed, then there must be some working versions
before new bugs are (re)introduced.

No, for a number of reasons.

First of all, there's *plenty* of *objective* evidence out there with
*specific data* showing that VueScan corruption is a given. And you
know this. You've been around this group long enough. A simple glimpse
of the archives is all the proof one needs.

As you yourself indicate by using the term "(re)introduced" the
so-called "fixing" of VueScan bugs is in reality *temporary masking*
of VueScan bugs. There are two indicators of this: One, the fact that
a "fixed" version usually corrupts the data even more in order to try
and *mask* the bug. A simple examination of data shows this clearly.
That's *not* a fix! Two, some (many!) VueScan bugs never seem to go
away but just "hide" for a version or two and then come back. That's
*not* a fix, either!

Finally, given such a "fluid" state of affairs and - more importantly
- the *massive* amount of VueScan bugs, some actual other caused by
faulty so-called UI confusion, etc, and you have more chance of
winning the lottery than getting uncontaminated VueScan output.

I only bother with raw scans on long nightime exposures, where it's not
helpful to do any semi-automated color adjustments in the scan program,
and I need every last bit of unmutiliated data (efficient workflow be
damned.) Your mileage will vary.

Indeed. Which brings us back to personal decisions we make. And that's
outside of the scope I write about. I may lay out some options but, in
the end, those are personal decisions/choices.

However, that has nothing to do with objective facts which lead to
only one conclusion: VueScan is far too buggy and unreliable to be
considered anything but amateur bungling. Nevertheless, it may suit
some people but they should not be under the misapprehension they are
getting "quality".

Have fun.

I always do! You too!

Don.

 

[Don]

Sorry to breach etiquette and repost the above in its entirety, but it shows
that there is still some intelligent life out there among the morons with
more time than brains.

Point(s) taken, and I'll pursue the RAW option to see what is coming out
without any decision-making by the scanning software. I always shoot my
digital in RAW, and don't need any further convincing that this is indeed
the way to go.

Thanks again.

You're most welcome. Glad to be of help.

Don.

 

[Roger S.]

The post I made wasn't really intended for Don, but rather a second
opinion on the usefulness of Vuescan, but here are replies to Don's
comments.

The profile I made for the Canon FS4000US with Scarse requires linear
gamma raw files (then apply the profile in Photoshop and convert to the
working space). It has limited usefulness as it's for Fuji chromes
only, but can be helpful at times. It only works for a given exposure.

I generally use the matrix profile feature in Vuescan which yields
pretty good color, requiring little or no subsequent editing (beyond
curves to bring up the shadows). There's no equivalent feature in
Canon's stock software. You get whatever colors it gives you and then
you get to spend time repairing them in Photoshop.

Vuescan exposure with this scanner is easy to figure out- a stopwatch
confirms confirms the scanner only has 6 exposure "speeds." If the
scan's taking too long, something is wrong. This hasn't been an issue.
If you set the exposure level too high, it defaults to the highest
exposure (6). This is predictable and easy to understand.

Canon uses an IR algorithm called FARE that is proprietary and distinct
from ICE. It leaves some strange digital noise-like artifacts on
negatives but works well on positives. As I mostly scan negatives,
that's not helpful. Vuescan's IR cleaning at worst doesn't do much and
leaves little dots around the dirt. That beats the large patches of
noise you get in FARE, as I can clone out the worst bits of dirt, but
not texture destroying patches of noise. At the moment, with IR
exposure bumped up in Vuescan from 1 to 3 (optimal level found through
time-consuming experimentation), the cleaning works pretty well,
removing obvious dirt without artifacts. I'd rather use ICE, but there
is no way to.

 

[Andre C]

Why do Don and Ron bicker so much? Are they married and having some
form of on-line tiff?

It has been fun reading your senseless ramblings for a while now but
can you not take this to a relationship counsellor or a lawyer?

 

[Steven Saunderson]

Why do Don and Ron bicker so much? Are they married and having some
form of on-line tiff?

Dialup Donny has been bagging Vuescan ever since he had an online tiff
with King Eddy. Ed produces software such as Vuescan while Don can't
produce software (DonnyScan is still vapourware) or even an image.

Just imagine the market for DonnyScan once he has eliminated Vuescan.
There might be some sensible commercial alterior motives behind his
apparently unreasonable behaviour.

 

[Don]

The profile I made for the Canon FS4000US with Scarse requires linear
gamma raw files (then apply the profile in Photoshop and convert to the
working space). It has limited usefulness as it's for Fuji chromes
only, but can be helpful at times. It only works for a given exposure.

The point is you have decided on a workflow without paying attention
to the software. Obviously that's not going to work but you can't
blame the software for the choices you made!

In other words, make a profile which does not require linear gamma raw
files and then there is no problem.

I generally use the matrix profile feature in Vuescan which yields
pretty good color, requiring little or no subsequent editing (beyond
curves to bring up the shadows). There's no equivalent feature in
Canon's stock software. You get whatever colors it gives you and then
you get to spend time repairing them in Photoshop.

For starters, VueScan's matrix color profiles are simplistic and of
very limited (if any?) use. Second, as I explained last time, the use
of scanner profiles in the first place (as opposed to monitor or
printer profiles) is of very limited use to start with because they
often do more damage than good.

So, once again, you're doing things with limited usefulness (like the
specific Fuji profiles above) and then blaming the software for your
choices.

Vuescan exposure with this scanner is easy to figure out- a stopwatch
confirms confirms the scanner only has 6 exposure "speeds." If the
scan's taking too long, something is wrong. This hasn't been an issue.
If you set the exposure level too high, it defaults to the highest
exposure (6). This is predictable and easy to understand.

Not knowing Canon as I mentioned before I can't comment on the
specifics. I can only say that measuring with a stopwatch is *not*
reliable testing. That's no different than looking at an image and
going: "Nah, it's good enough." Again, that's not reliable testing,
you need to look at the (16-bit!) histogram, etc.

The bottom line is you really don't know what is going on and what
damage VueScan does.

Canon uses an IR algorithm called FARE that is proprietary and distinct
from ICE. It leaves some strange digital noise-like artifacts on
negatives but works well on positives.

Again, not knowing Canon I can only make general comments. It's very
odd that software would work fine on positives but not on negatives.
Possible, yes, but quite improbable. So, my suspicion would be user
error. But, again, I just don't know enough about the environment.

As I mostly scan negatives,
that's not helpful. Vuescan's IR cleaning at worst doesn't do much and
leaves little dots around the dirt.

Depends on the level. If you choose higher level VueScan does
considerable damage which is even visible to the naked eye. That's
what those recurring IR VueScan bugs are all about.

That beats the large patches of
noise you get in FARE, as I can clone out the worst bits of dirt, but
not texture destroying patches of noise. At the moment, with IR
exposure bumped up in Vuescan from 1 to 3 (optimal level found through
time-consuming experimentation), the cleaning works pretty well,
removing obvious dirt without artifacts. I'd rather use ICE, but there
is no way to.

For what it's worth, and in spite of how wonderful ICE is, in the end
I for one decided against it even on film compatible with ICE. I
started with Kodachromes and ICE doesn't work with them so I got used
to manual removal of dust and scratches. And then once I progressed on
to Ektachromes I was spoiled by the sharpness of non-ICE Nikon scans
(warped film notwithstanding) that I decided to continue with manual
dust and scratches removal.

But that, again, is a personal choice.

Don.

 

[Randy Howard]

Steven Saunderson wrote

Just imagine the market for DonnyScan once he has eliminated Vuescan.

If somebody wants to eliminate a product, they've got to do more
than post on Usenet. If it was that easy, Windows would have
disappeared a decade ago.

There might be some sensible commercial alterior motives behind his
apparently unreasonable behaviour.

What is an "alterior" motive?

 

[Steven Saunderson]

What is an "alterior" motive?

Arrgh, how embarrassing. I'm sure there must be a problem with my
keyboard. At least I've now learnt to spell "ulterior".

Thanks

 

[Randy Howard]

Steven Saunderson wrote

Arrgh, how embarrassing. I'm sure there must be a problem with my
keyboard. At least I've now learnt to spell "ulterior".

Thanks

Muybe someone chunged the urrungement of the the 'u' and 'a'
keys.

 

[Steven Saunderson]

Muybe someone chunged the urrungement of the the 'u' and 'a'
keys.

Yes, I definitely need a DWIM keyboard. After consulting with the
powers that be (my CC balance) I have now bought a licence for Agent so
I can use its spelling checker. I like Agent and it certainly has given
me good value over the years.

 

[Roger S.]

Vuescan exposure with this scanner is easy to figure out- a stopwatch

confirms confirms the scanner only has 6 exposure "speeds." If the
scan's taking too long, something is wrong. This hasn't been an issue.
If you set the exposure level too high, it defaults to the highest
exposure (6). This is predictable and easy to understand.

Don: "Not knowing Canon as I mentioned before I can't comment on the
specifics. I can only say that measuring with a stopwatch is *not*
reliable testing. That's no different than looking at an image and
going: "Nah, it's good enough." Again, that's not reliable testing,
you need to look at the (16-bit!) histogram, etc. "

Of course, it's patently obvious if the exposure weren't set correctly
as you'd have blown highlights in slides. You don't need a 16 bit
histogram to see that. The actual exposures can be easily confirmed in
a program like Photomatix which shows the image histograms before you
combine scans at different exposure levels. The histograms clearly
progress rightwards as you increase exposure stops. In this case, VS
is working as it is expected to. The stopwatch merely confirms it. If
you set exposure to 100, you only get the exposure of 6. The
histograms and time taken are identical.

 

[theo]

Why do Don and Ron bicker so much?

It seems that Don is this NG's self-annointed self-appointed Ancient
Mariner (cf Coleridge, 1797 and 1817) proclaiming "beware!", with some
affect of the crazed comedian -"...slowly I turned, step by step, inch by
inch..." should one utter the shibbolith "vuescan". Only proles, having
no sense of appreciation for objective evaluation, give $ to Hamrick, and
"...fools and their money...". Don will again in future threads with
self-righteous pride parenthetically pontificate upon his self-discipline
in refusing to respond to all his critics' categorical ad hominem attacks,
and all criticisms are of course ad hominem. But I have not tripped
across any links to his scan/editing protocols nor to ASP's authors lists
for all the code he has written nor any you-show-yours-and-I'll-show-mine
graphics prowess. Won't someone show me the way toward his enlightened
processes and/or product?
=======
Pessimists remain morose precisely because they are so right too often.