Vuescan - new features

  • Thread starter Thread starter Robert Feinman
  • Start date Start date
Evo2Me ([email protected]) wrote in
There's no sense in arguing with Don - he reacts selectively and
ignores
the arguments he cannot refute.

Since when can Don refute anything - or want to? When I put him into
my killfile I did not because I don't like dissenting opinions but
because he's nothing more than a zealot [religious allusion
intentional] on his war path against Mr Ed Hamrick and his program.

I beg to disagree - he may be that (I don't care) but he _is_ more.
Which you of course wouldn't notice if you have killfiled him. I find
his posts valuable when he's not being a zealot, which most of the time
he isn't. ;-)
 
Yes, you can set points directly (i.e. blind) in the curve but that's
a very blunt instrument. Doing that means making assumptions about the
image. For example, using the "standard" S curve (64,128,192) assumes
where the shadows, midtones and highlights are.

However, even in that case, basing this "blind" setting on a gamma 2.2
image while the curve is applied to a gamma 1.0 image doesn't get
around that root problem.

It is the curve that matters, not where the control points are. Selecting
a curve is an interactive/iterative process. Getting some idea where the
'bad' parts of the image fall on the curve is useful, not not required.

For proper control, it is probably best of the curve window is based on LAB,
the input is based on XYZ and the display is based on the monitor profile
(typically around gamma 2.2).
It wouldn't be the first time that Vuescan fails to do the simplest of
things which is precisely why it can be justifiably described as buggy
and unreliable.

That depends. If new features don't work the first time, send a detailed
bug report to the author and it will probably be fixed.

I don't have a Vuescan license because I can usually live with the limitation
of the vendor supplied software.

But the trail versions I tried were never buggy in my experience. Just
not worth the money.
Which is precisely the root of the problem, apparently.

That is not the root problem. It is supposed to work like that.
I notice quotes around 'cannot' which means you know what I'm going to
say (so just ignore the next bit) but for the benefit of kibitzers:

Actually, one can. That's what linear editing is all about. One simply
recalibrates the monitor to gamma 1.0 as well and then the image does
not appear "dark and contrasty".

The 'cannot' is there for two reasons:
1) of course you display a gamma 1.0 image directly on a gamma 2.2 monitor:
it just won't look very good.
2) as far as I know all my graphics cards have 8-bit/ch D/A converters and
8-bit/ch is not enough for gamma 1.0. Of course there is the additional
problem that in general monitors don't support gamma 1.0 anyhow.

(Trying to use 8-bit/ch LUTs to convert from gamma 1.0 to 2.2 is bound to make
things even worse).
But even in that case looking at one image while editing another (in
the current context) is just asking for trouble. But such convoluted
contraptions are one of the major trademarks of Vuescan and why it is
so buggy and unreliable.

Looking at one image while editing another is exactly what my XYZ library
does. There is no magic there. (Of course, if you select in Photoshop
any color space other the monitor color space, you end up with the same
thing. Typically not with luminance, but with saturation).
 
Don said:
Exactly! And even if you didn't Photoshop would do the necessary
conversions both ways.

Well, my monitor is set to 1.8 and I think that Photoshop indeed handles
the conversions correctly. I don't see what makes you conclude that
VueScan doesn't. I don't use curves in VueScan, but from what I
understand you can only set anchor points at the bins corresponding to
25% and 75% at gamma 1.0, and while you adjust the output values, you
see the effects of your adjustment in gamma 2.2 (or whatever output
gamma you've set). This may not be the most straightforward approach but
as long as it's WYSIWYG I think it's OK.
 
SNIP
I don't use curves in VueScan, but from what I understand you can
only set anchor points at the bins corresponding to 25% and 75% at
gamma 1.0,

Not exactly. Given the frequent user request for a type of S-curves
control, Ed Hamrick chose to implement it as a fixed midpoint, and two
sliders for shadow and highlight adjustment (sliders default to 25%
and 75% of WP-BP range).

That's a good start, although I personally would welcome the addition
of a mid/pivot point control. One could also debate whether the
defaults could/should have been set somewhere else (which could be set
to any user preference if the pivot point was user adjustable).
and while you adjust the output values, you see the effects of your
adjustment in gamma 2.2 (or whatever output gamma you've set).

And after e.g. WP, BP, color management.
This may not be the most straightforward approach but as long as
it's WYSIWYG I think it's OK.

Yes, it is WYSIWYG, so output is quite predictable.

Bart
 
Since I'm the one who started this thread I'll add a few notes.

1. Ed does not read the usenet group anymore (or at least he does
not participate).

2. He has always responded appropriately to my bug reports and/or
suggestions. Last month I suggested he add a feature that allows
for switching between images when scanning multiples at once
without needing to go back to the input panel. He added this quite
promptly.

3. Part of the problem is that people have different expectations
from software. Those who just want the image captured with a minimum
of post-processing find Vuescan a good choice. Those who want all the
bells and whistles may prefer Silverfast. Having a choice is a "good
thing." If you don't like one, then don't use it.

4. The troll problem seems incapable of being fixed. Kill filing
specific posters has certainly made my group reading much more
efficient. Missing specific posters also points out how unproductive
most of the responses are. The argument that a troll needs to be
debated or a newcomer may get bad advice may seem noble, but is
probably ineffective. If one feels he must alert a questioner as
to the credibility of a troll then a simple statement like:
"So and so's advice is usually not respected by knowledgeable members
of this group and I suggest you look at other viewpoints as well."
Then leave it at that. Do I think the impassioned will take my advice?
No.

5. Since I posted my announcement of the new version, there have been
a couple of further upgrades. Read the what's new on the Vuescan site
to see if they improvements affect you. Overall the support for this
product seems above average. See what sort of response you get to a
Microsoft bug report. Many companies now charge $20 for you to talk to
a tech even to report a defect in their product or service.

Finally, I think the drop off in discussions here reflects the overall
decline in the use of film and thus in the number of people buying or
using scanners.
 
Well, my monitor is set to 1.8 and I think that Photoshop indeed handles
the conversions correctly.

I know it does because I played around with linear editing a bit. Both
Photoshop and NikonScan can work in gamma 1.0. I set my monitor
temporarily to 1.0 and while the desktop looked "funny" the images in
both Photoshop and NikonScan were clearly interpreted correctly.
I don't see what makes you conclude that
VueScan doesn't.

Two things: One is Bart's post and the other is that Vuescan is
traditionally (read: track record) prone to doing things like that.
I don't use curves in VueScan, but from what I
understand you can only set anchor points at the bins corresponding to
25% and 75% at gamma 1.0, and while you adjust the output values, you
see the effects of your adjustment in gamma 2.2 (or whatever output
gamma you've set). This may not be the most straightforward approach but
as long as it's WYSIWYG I think it's OK.

Yes, it is important to see the result but if indeed there are only 2
fixed points that can hardly be described as curves. Not to mention
that they are applied to the "invisible" gamma 1.0 image.

BTW, if that's the case, it's a prime example of what I just refer to
above i.e. Vuescan doing things in a convoluted and non-standard way
for no reason, other than it was probably the most convenient for the
author which is not exactly a sign of professionalism.

Don.
 
Not exactly. Given the frequent user request for a type of S-curves
control, Ed Hamrick chose to implement it as a fixed midpoint, and two
sliders for shadow and highlight adjustment (sliders default to 25%
and 75% of WP-BP range).

That's a good start, although I personally would welcome the addition
of a mid/pivot point control. One could also debate whether the
defaults could/should have been set somewhere else (which could be set
to any user preference if the pivot point was user adjustable).

If that's the case, fixed points make it virtually useless. It's yet
another typical convoluted Vuescan half-measure masquerading as
something else.
And after e.g. WP, BP, color management.

Oh! That too? So, essentially the settings are virtually random?
Yes, it is WYSIWYG, so output is quite predictable.

No, it's not! You may be *seeing* the result but the *settings* are
certainly not predictable by any stretch of the imagination given the
multitude of convoluted layers between the image and the end result
(gamma, WP, BP, CM and who knows what else... oh yeah, bugs!).

Don.
 

Oh, OK. If you say so without any supporting evidence it must be true.

That clearly demonstrates your unique concept of "objectivity".
Right. So being objective means arguing your point based on someone elses
experience.

In addition to difficulties with being objective you also seem to
struggle with the concept of "reference"?

Look at any scientific document and you will see a list of references.

According to you all that's "gossip" and such a document author is
using "someone else's experience".

Newsflash: That's how science works! You identify your sources clearly
and provide as much information as possible.

*Not* to have clearly identified the source would have been hearsay.
hear·say
n.

1. Unverified information heard or received from another; rumor. 2.
Law. Evidence based on the reports of others rather than the personal
knowledge of a witness and therefore generally not admissible as
testimony.

So, how exactly have you verified your source?

By *identifying* both the source and the context. See above.

Another hint: If you want to play semantics and etymology do note the
use vague "from another" in your definition clearly implying an
anonymous source.

Don.
 
Maybe so. I wouldn't know as I also have him kill-filed, but judging
by his history

And yet you just can't stop obsessing with comments!?

For someone who takes every opportunity to loudly proclaim killfiling
someone you spend an awful lot of time commenting which makes your
protestations about killfiling sound very hollow, indeed.
I can only hope you and others can make a distinction between fact and
fiction. Kill-filing still seems a better option, to me anyway.

As always, when it comes to Vuescan you stop being rational and keep
contradicting yourself. Does that sound harsh? Maybe, but unlike you I
base my statements on fact:

As I expected your reaction to my message, I took the courtesy to
temporarily let it show in my newsreader

Riiight... "Temporarily!" Nudge-nudge, wink-wink...

Don.
 
I find
his posts valuable when he's not being a zealot, which most of the time
he isn't. ;-)

Only "most of the time"? ;o)

That means I'm not a very good zealot and I have to try harder! ;o)

Don.
 
It is the curve that matters, not where the control points are.

But the control points determine the curve!
Selecting
a curve is an interactive/iterative process. Getting some idea where the
'bad' parts of the image fall on the curve is useful, not not required.

That's where we disagree. Yes, it is possible to use a generic S curve
for example but, as I mention, for anyone who actually wants to know
what they're doing it's essential to identify the parts of the image
one wants to change and, conversely, parts of the image one doesn't
want to disturb.
That depends. If new features don't work the first time, send a detailed
bug report to the author and it will probably be fixed.

I don't use Vuescan, but judging by reports from other users author's
reactions fall into two broad categories if he can't fix the bug:
He explodes with abuse or tells the user they're "blacklisted" -
whatever that means.
I don't have a Vuescan license because I can usually live with the limitation
of the vendor supplied software.

But the trail versions I tried were never buggy in my experience. Just
not worth the money.

That too. But it depends how thorough your tests were. In any case,
Vuescan's track record of all the bugs is in the public domain.

As I keep saying, for a highly compressed web JPG even Vuescan will do
and I have even recommended it to people who have such low
requirements. But for anything else it's pretty useless.
That is not the root problem. It is supposed to work like that.

It may be supposed to work like that, but that's not an excuse. A
design flaw is still a design flaw, even if it's implemented as
intended which in case of Vuescan is always a question mark because of
all the bugs.
The 'cannot' is there for two reasons:
1) of course you display a gamma 1.0 image directly on a gamma 2.2 monitor:
it just won't look very good.

Indeed! As I mention in a parallel message I tried it once and the
posterization made it virtually unusable. But it did "work".
2) as far as I know all my graphics cards have 8-bit/ch D/A converters and
8-bit/ch is not enough for gamma 1.0. Of course there is the additional
problem that in general monitors don't support gamma 1.0 anyhow.

Exactly! Hardware limitations also cause posterization at gamma 1.0.
Even at gamma 2.2 the limitations of 8-bit color are potentially a
problem when working with 16-bit images, not to mention HDR.
(Trying to use 8-bit/ch LUTs to convert from gamma 1.0 to 2.2 is bound to make
things even worse).

Again, I couldn't agree more! Which is yet another reason to avoid
unnecessary conversions and "guesswork" but try to be as exact as
possible. Which is why I would not describe a fixed 2-point S
correction as "Curves".
Looking at one image while editing another is exactly what my XYZ library
does. There is no magic there. (Of course, if you select in Photoshop
any color space other the monitor color space, you end up with the same
thing. Typically not with luminance, but with saturation).

Yes, but that's different which is why I noted "in the current
context".

It's quite common to look at one image while working on another and I
can list a number of such cases when such an approach makes sense. But
the problem here is it does not.

Anyway, it's all academic because apparently in case of Vuescan it's
even worse. There is no image to look at all! The "curves" can only be
set "blind" using *2 fixed* control points and are applied *after* a
whole bunch of other processing.

Don.
 
Marjolein Katsma said:
But it robs you of any ground on which to conclude whether he's a only
zealot or more than that.

Rest assured that Bart, me, and a lot of people have had ample
opportunity to reach our conclusions on this matter.

Ralf
 
Bart van der Wolf ([email protected]) wrote in [email protected]:


But it robs you of any ground on which to conclude whether he's a only
zealot or more than that.

That's because Bart made up his mind based on emotion and will not
allow objective facts to change it.

Why else continue obsessing so intensely with comments to messages
from someone he so loudly proclaims to filter? It just doesn't make
sense.

It's really a shame, because I do value Bart's contributions even
though I don't agree with his conclusions.

Don.
 
Robert Feinman said:
Since I'm the one who started this thread I'll add a few notes.

1. Ed does not read the usenet group anymore (or at least he
does not participate).

I assume he reads, but doesn't participate (saves him a lot of phony
"bug" reports). Reading Usenet and Web posts about one's creation
could provide useful user feedback. Like with scanning, the real issue
is to discriminate between "signal and noise".
2. He has always responded appropriately to my bug reports
and/or suggestions.

Same with me (with one exeption on the DSE-5400 Mk 1 D-max, for me
anyways). Depending on time zone and informed description of
issue/request, issues may get tackled in hours, rather than annually,
and at zero cost. Resolution of some issues may (for a one man
enterprise) depend on popularity / availability / manufacturer and
urgency (seems good longer term business sense).

The addition of a type of curves control is an example, asked for by
users, it got implemented after solving more pressing issues (afterall
curves-control duplicates already available postprocessing
functionality).

SNIP
3. Part of the problem is that people have different expectations
from software.

Well, on Usenet the trolls are the real "issue", somewhat 'solvable'
by kill-filing them, but not a real solution for newbies (= by the way
not a value judgement).
Those who just want the image captured with a minimum of
post-processing find Vuescan a good choice. Those who
want all the bells and whistles may prefer Silverfast.

IMHO, VueScan offers more "bell and wistle" control as well.
Having a choice is a "good thing." If you don't like one, then
don't use it.

Competition is great (poor 'marketing' isn't)!
4. The troll problem seems incapable of being fixed.

"Seems" may be the key word.
Kill filing specific posters has certainly made my group reading
much more efficient.

Hear, hear! Although it's a bit like (covering one's ears) I don't
hear you (but then, in fact, little is lost)!

SNIP
Overall the support for this product seems above average.

Yes, I agree that examples of similar support are rare, despite futile
troll attempts.

SNIP
Finally, I think the drop off in discussions here reflects the
overall
decline in the use of film and thus in the number of people buying
or
using scanners.

Yes, unfortunately for those who joined "late". I intend to
participate if not 'discouraged' by more trolls (and even that might
increase my determination of exposing them for what they are). This
group has turned out to be a great source of reflection/information
for me, so some reciprocity is called for IMHO.

Bart
 
Evo2Me ([email protected]) wrote in
There's no sense in arguing with Don - he reacts selectively and
ignores
the arguments he cannot refute.

Since when can Don refute anything - or want to? When I put him into
my killfile I did not because I don't like dissenting opinions but
because he's nothing more than a zealot [religious allusion
intentional] on his war path against Mr Ed Hamrick and his program.

I beg to disagree - he may be that (I don't care) but he _is_ more.
Which you of course wouldn't notice if you have killfiled him. I find
his posts valuable when he's not being a zealot, which most of the time
he isn't. ;-)

I like to hear differing opinions. Well presented they give me a
foundation on which to judge the topic, but when the discussion turns
to a rant, or crusade. When it becomes a crusade and discussion
becomes futile it's time to either ignore the thread, ignore the
poster, or use the "kill file" button..

It's unfortunate, but any person can make a 100 good, informative
posts, but it only takes one that goes off on a tangent to become
ignored. When the tangent turns into a rant, or that crusade, many
reach for the "kill file" button and never hear any of the possible
good stuff again.

Newsgroups can be a large pool of knowledge are a great place to
listen to and voice opinions to a rather large group. but get carried
away and the audience gets smaller in a hurry.


Roger Halstead (K8RI & ARRL life member)
(N833R, S# CD-2 Worlds oldest Debonair)
www.rogerhalstead.com
 
[...]If one feels he must alert a questioner as
to the credibility of a troll then a simple statement like:
"So and so's advice is usually not respected by knowledgeable members
of this group and I suggest you look at other viewpoints as well."

Although pragmatically I concur with you, there is a problem: why
should anybody who doesn't know the people and discussions in a
newsgroup believe you but not me?

Another, broader problem, is that even if the majority finds one
specific poster to be a nuisance it doesn't necessarily follow that
this poster is wrong. Democracy is nice but doesn't tell you anything
about the truth. Which is exactly why it needs discussion and
refutation.

Otherwise I am totally with you, same experiences concerning Ed
Hamrick and Vuescan.
 
I know it does
Two things: One is Bart's post and the other is that Vuescan is
traditionally (read: track record) prone to doing things like that.

I don't know which post you are referring to in which art points out
that VueScan does not handle the conversions correctly.
Your second argument is not proof. Even if you come up with a 'track
record' of 1,000 pages it still isn't.
Yes, it is important to see the result but if indeed there are only 2
fixed points that can hardly be described as curves. Not to mention
that they are applied to the "invisible" gamma 1.0 image.

I think it's possible to set VueScan's output space to gamma 1.0 and
make the 'invisible image' visible. If you want to.
BTW, if that's the case, it's a prime example of what I just refer to
above i.e. Vuescan doing things in a convoluted and non-standard way
for no reason, other than it was probably the most convenient for the
author which is not exactly a sign of professionalism.

There is some truth in that. Perhaps the way VueScan's "curves" work
could perhaps better be compared to Photoshop's contrast adjustment.
Compared to that it offers more control but it's certainly inferior to
Photoshop's curves adjustment. That's why I don't use it, but it may be
useful for GIMP users who cannot do anything in 16 bits after scanning.

On the other hand, what is a 'standard'? Is there an ISO, ANSI or DIN
standard for curves adjustments? I think the way Photoshop, PaintShop
Pro and Picture Window Pro do it, is a de-facto standard just like the
'standards' Microsoft puts us up with. If somebody comes up with a
better solution for the same problem, he or she could have an advantage
on the market. If Ed Hamrick has an inferior solution, he may lose
market share - or people just don't use his curves adjustment feature.
 
I don't know which post you are referring to in which art points out
that VueScan does not handle the conversions correctly.

This one:

--- start ---
The curves control feature is implemented a bit differently compared
to most photo editors. The adjustment is applied to linear gamma data
after setting black/whitepoints, but before Gamma adjustment. So don't
expect to use the same settings as one would use in Photoshop. The
Preview/Scan tab will show the effect after all (including gamma and
colorspace) adjustments, so go by that.
--- end ---

Even though Bart agonizes trying to minimize the problems by using
vague and circumspect language, the truth still comes through if one
reads carefully.
Your second argument is not proof. Even if you come up with a 'track
record' of 1,000 pages it still isn't.

It's certainly no proof positive in the absolute sense, but you can't
just casually dismiss a known and appalling track record based on
reliable sources.

Sure, you can give an unknown product such a benefit of the doubt, but
after a product repeatedly demonstrates its unreliability it would be
irrational to ignore its track record.
I think it's possible to set VueScan's output space to gamma 1.0 and
make the 'invisible image' visible. If you want to.

I don't think you can change the gamma space but we are both
speculation about something which is not important.

The important thing is that this Vuescan contraption is anything by
"Curves". In addition to all the "oddities" Bart has outlined, having
only two fixed points and working totally blind makes it basically
useless.
There is some truth in that. Perhaps the way VueScan's "curves" work
could perhaps better be compared to Photoshop's contrast adjustment.

I think you're right about the intention because the "S" curve is
really a contrast adjustment.
Compared to that it offers more control but it's certainly inferior to
Photoshop's curves adjustment. That's why I don't use it, but it may be
useful for GIMP users who cannot do anything in 16 bits after scanning.

That's a stretch (because it addresses a very narrow subset) and I'm
sure the author did not even consider that.

Instead he wanted to be able to (falsely!) advertise Vuescan as having
Curves.
On the other hand, what is a 'standard'?

The Vuescan "curves" contraption has nothing to do with standards. The
standards I refer to above are things like using EV for exposure.

This is about common sense and logic (the "convoluted" part above).
It's just bad programming (i.e. a bug) to examine one image and apply
the settings to another (in this context).

As it turns out, the truth is even worse because the settings are
fixed and applied blindly after a lot of other pre-processing.

Don.
 
Back
Top