VueScan and really dark images
Marcel-Jan Krijgsman
Recenly I did some experiments with long exposure photographs of the
nightsky (20, 30, 60 seconds). Just looking at them with the eye, I was
quite happy with what I saw.
I wanted to scan them with my Benq Scanwit 2740S and VueScan 7.5.54, but
I could tell that there would be difficulties, because, well, they are
really dark. And as expected VueScan 'helped' me by brightening up the
picture tremendously. But that wasn't wat I was looking for. So first I
put the Color balance (under the Color menu) on 'None' and then on
'Night'. 'None' was better, but still it was too bright. I fiddled with
some other options like 'Slide curve'.
I'm very interested to hear suggestions how to improve scanning these
nightsky images.
Thanks,
Marcel-Jan Krijgsman
Tom Harrison
Were these slides or negatives? I have a similar film scanner (Minolta Scan
Dual II) with similar specs and think you may be seeing one or both of the
following:
1) If slides, the slides are going to be almost pure black. This is a very
challenging task for any scanner since the scanner needs to be able to shine
a strong enough light through the film to get a signal to the CCD. Your
scanner has a modest dynamic range (3.2 according to specs, out of a maximum
of 4) which is where this comes into play; the larger the dynamic range, the
more of the full spectrum of grays from white to black can be recorded. So
you can either underexpose one area to get the right exposure on another, or
vice versa. In the absence of a strong enough signal (underexposure) you
may see electronic noise, apparent as speckles (pixels) that should be pure
black but are some other color. If you overexpose, where there *is* light
in the image, the scanner may "bloom" where the charge from one pixel spills
to another, causing it to be overexposed.
VueScan has the ability to increase exposure manually: Device|Lock Exposure
will reveal the RGB exposure setting. Check the user guide under Advanced
Workflow suggestions for a good procedure to set that. Further, you may
also have luck using "Long exposure pass" which essentially takes a normal
exposure and then an overexposure and merges the best of the two ...
although this can fail at the edges, so your results may or may not be any
good in this case.
Negatives are different: the emulsion is thinner and usually less problem
for scanners, and furthermore you should have an almost purely clear frame.
In this case, use the advanced workflow suggestions (in user guide) to lock
the exposure and film base color.
2) Whether slides or negatives, you may also see VueScan trying to do what
it does well under normal conditions, which is to provide a white-balance of
the scene. This is fine in daylight, but will tremendously amplify
electronic noise in the scene you are working on by lightening everything.
Depending on what you are scanning, you'll probably want to try either a
Night, Neutral, or None for color balance. You may also want to try this
with the Curves feature set to "Log (Dark)". You can get a pretty good idea
of what the scanner actually saw by looking at the scan result with color
balance set to None, this and the histogram may help guide you in how to
adjust for this lighting condition.
Keep in mind that the two problems can interact: a very noisy scan (perhaps
resulting from underexposure, very dense slide film, or a lower quality
scanner) will be highlighted by any VueScan adjustments that make things
brighter. Get the exposure right first using None for color balance, then
get the color balance right.
Hope this helps.
Tom
"Marcel-Jan Krijgsman" <m.kri...@tiscali.nl> wrote in message
news:3DBD575D...@tiscali.nl...
Gary Hundt
>
> Recenly I did some experiments with long exposure photographs of the
> nightsky (20, 30, 60 seconds). Just looking at them with the eye, I was
> quite happy with what I saw.
By "looking at them with the eye" do you mean looking at the prints
you got back from the processor?
>
> I wanted to scan them with my Benq Scanwit 2740S and VueScan 7.5.54, but
> I could tell that there would be difficulties, because, well, they are
> really dark. And as expected VueScan 'helped' me by brightening up the
> picture tremendously. But that wasn't wat I was looking for. So first I
> put the Color balance (under the Color menu) on 'None' and then on
> 'Night'. 'None' was better, but still it was too bright. I fiddled with
> some other options like 'Slide curve'.
>
> I'm very interested to hear suggestions how to improve scanning these
> nightsky images.
>
> Thanks,
> Marcel-Jan Krijgsman
Try drastically increasing the Black Point %. (Also the white point,
probably.) This will make them look more like the prints you probably
got back from the processor.
Note that when taking such long exposures, you may be allowing much
more light onto the film than you can pick up with your own eyes even
with pupils wide open. Your long exposures may look more like the
scene would look to a nochturnal animal!
Hmm... it would be worth figuring out for a given film ISO, what the
exposure time corresponding to the human eye's maximum light
sensitivity is. Then, if you just want to record what you see, you'd
just use this exposure time without having to bracket excessively
(unless you want also to maintain detail in regions lit by street
lights, moon, etc - then you'd have to expose for these - just like
when you shift your eyes from the shadows to a lit area, your pupil
closes down accordingly.)
-Gary
Bart van der Wolf
> I wanted to scan them with my Benq Scanwit 2740S and VueScan 7.5.54, but
> I could tell that there would be difficulties, because, well, they are
> really dark. And as expected VueScan 'helped' me by brightening up the
> picture tremendously. But that wasn't wat I was looking for. So first I
> put the Color balance (under the Color menu) on 'None' and then on
> 'Night'. 'None' was better, but still it was too bright. I fiddled with
> some other options like 'Slide curve'.
It's not clear to me whether you're using slide film or print film. However,
you probably need to adjust the Color|White point to 0.001% or 0%, and then
adjust the Brightness. You probably don't want to change the 0% Black point
setting, but that depends on the image (stars, nebulae, planets of perhaps
moonlit cloud formations or cityscapes) and the resulting histogram. The
Neutral color balancing will give you most flexibility for accurate colors.
Knowing the type of film (reversal or negative) would help in giving a more
to the point suggestion for improvement (something along the lines of the
Advanced Workflow Suggestions).
Bart
Sebastian Voitzsch
> I could tell that there would be difficulties, because, well, they are
> really dark. And as expected VueScan 'helped' me by brightening up the
> picture tremendously. But that wasn't wat I was looking for. So first I
> put the Color balance (under the Color menu) on 'None' and then on
> 'Night'. 'None' was better, but still it was too bright. I fiddled with
> some other options like 'Slide curve'.
>
The Scanwit in general seems to be set up for negatives. This causes every
slide, not only night skys, to appear very dark. I would do a RAW scan in
vuescan (i.e., without any processing in vuescan), take the pic to a photo
editor and adjust the curve. Make it look like an arc (from bottom left to
upper right, ofcourse ;O)) as this gives the best result IMHO. A nice
"howto" can be found at
http://www.photoscientia.co.uk/Review.htm#Usage
(short description in the review, more on the links on bottom of the page).
HTH
Sebastian
Marcel-Jan Krijgsman
I meaned while looking with the naked eye at the slide. Yes, slides, sorry I forgot to mention that.
Okay, you mention Color | Black point so I did some experimenting with that value. I had Color Balance
set to 'None' and then you don't get to choose the Black point. So now I set Color | Color Balance to
'White Balance' and set Black point (%) to 1 and White point (%) to 0. And indeed, it comes out much
darker. Probably Color Balance on 'Night' would be more sensible, but I like what I'm seeing. Some more
experimenting will be required.
Thanks for pointing that out.
Marcel-Jan Krijgsman
Marcel-Jan Krijgsman
Thank you all so far for the tips so far. I'm gonna try most of it, until I
get the most acceptable result and all the tips seem worthwile to try,
possibly in combination.
(To bad this isn't a binaries group, otherwise I could have shown some
results.)
Thanks,
Marcel-Jan Krijgsman
Beaker
Marcel-Jan, do you have some webspace?
If so, you could upload your results and just post the link here. :O)
--
=8O(
Ian Cox
Sutton-in-Ashfield.
icq 116510696