B/W workflow to improve on Vuescan/Minolta scans?

[mendel_leisk]

Up till now, my black and white workflow has entailed using Vuescan

scan-from-disk with Vuescan produced raw files, setting image type as

B/W negative. My baseline settings were TMax400 with D76ci:.55, .02%

clipping of wp and bp and .9 bright. It works, but I have concerns.

For one, it bothers me having to clip to improve contrast. Also, I

suspect the Vuescan b/w profiles are applying a contrast reducing s-

curve. and if I then apply a contrast enhancing s-curve in Photoshop,

I'm correcting a correction.

 

Possibly as a consequence of the clipping and black and white

profiles, the quality seems to suffer. Grain seems accentuated,

smooth tone transitions seem "mealy". Also, sometimes Vuescan's

exposure calculation seems sometimes spot-on, other times it seems

designed to confound.

 

I'm now trying the following adjunct to the Vuescan workflow. It is

specific to Minolta scanners, I think, due to the specific Minolta

ICC profile:

 

1. Do the best I can with Vuescan.

 

2. Open the source Vuescan raw file in Photoshop CS, apply the "Scan

Dual2 (posi)" profile, then convert to regular profile (simple rgb in

my case). Invert the image. At this point I have a passable, albeit

pretty flat, magenta tinted positive image.

 

3. Open the Vuescan attempt.

 

4. With the Photoshop CS histogram display on, note the "mean"

histogram value of the Vuescan attempt.

 

5. Switch to the converted raw, in levels apply 0 clip per channel,

adjust the mid-range pointer (gamma setting) to approximate the

previous noted "mean" value, and apply.

 

6. Note, this section seems to rarely require any action, just the

occasional image: Compare the "std deviation" histogram value, of

Vuescan attempt vs converted Vuescan raw. Usually at this point, the

convert raw's "std deviation" value will be a little higher. If it's

not and the converted raw appears a little too flat, apply a slight

contrast raising s-curve. Also compare highlight and shadow detail.

The converted raw is usually a little more contrasty, but within

reason. If you feel the contrast has gone too far, apply a slight

contrast lowering s-curve. Alternately, use Photoshop CS's

highlight/shadow tool, with quite low settings.

 

7. Convert to greyscale using Image|Mode|Greyscale. This seems quite

satisfactory, for conversion of colour scans of black and white film,

to my eye.

 

8. Do a "Save As".

 

Comparing the Vuescan attempt to this raw file conversion, I've been

finding:

 

1. Absolutely no highlight blowout, coupled with an image that still

has more punch, and cleaner shadows.

 

2. Smoother, truer looking tone gradations.

 

3. Smoother grain in some tone ranges, at least equal in others.

 

Thanks in advance for any comments and feedback.

[og]

Mendel,<p>

 

Your new workflow is extremely interesting.<br>

I haven't experienced yet my Minolta Elite 5400 with Black & White... But for difficult Scans, I will probably scan Negatives as 'Positive', while maximizing exposure without clipping (to reduce noise), and work directly on the Raw 16-bits linear output (invert, auto-level with 0 clipping, then Gamma correction - I have yet to find the appropriate value).<p>

 

 

Somme comments:<br>

- You may consider a fixed value for gamma correction, instead of the ones proposed by Vuescan (based on you own likings).<br>

- Some people advice to work in color and keep only 1 channel (or mix RGB), to reduce noise. I would be interested to hear more about this...<p>

 

I would be very interested to see some comparison of your results.<br>

Thank you,<p>

 

Olivier

[mendel_leisk]

Hi Olivier,

 

If I gamma correct the Vuescan raws directly in levels, the histogram develops frond-like comb teeth (only way I can think to describe it). Conversly, the "assign profile/convert to profile" approach using (in my case with Minolta Dual II) "Scan Dual2 (posi input)" profile, followed by invert, gets the gamma in the ball park, recquiring only minor subsequent adjustment, without apparent degradation of the histogram. This was suggested a few months back in this thread:

 

http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=005V1d

 

Any Minolta scanner seems to install 2 icc profiles along these lines, in my case they reside in:

 

C:\WINDOWS\system32\Color

 

And are named:

 

MLTF2820Un.icc

MLTF2820Up.icc<--this one coresponds to Scan Dual2 (posi input) profile in ps, my case

 

Regarding your comments:

 

1. Choosing gamma is something I struggle with, perhaps because it is the most subjective element. The film I'm scanning is of uneven density, I would like to be able to apply a single correction (after the profile conversion process), but I don't think it's practical. The original contact sheets I made, years ago, show large variations in exposure. It would be interesting to understand how Vuescan makes it's decision. I would welcome any suggestions on objectifying the process.

 

2. In my case, I find the green channel "looks" best, deeper cleaner shadows, more pop. Still I find little difference between converting to greyscale vs saving just green. I think using channel mixer really comes to the fore when converting true colour scans to black and white.

 

Here are some examples. I'm not sure how to post all in one so will just string them out:

 

First is Vuescan output, overall picture, reduced, with my default settings, Tmax400/D76ci:.55/.02%wp&bp/.9bri

 

Second is raw conversion, attempting to match first's "mean" value in histogram, at the rgb stage. With conversion to greyscale this mean value drops, slightly. No additional curve has been applied.

 

Third and fourth are identical crops of one and 2 respectively.<div></div>

[mendel_leisk]

second<div></div>

[mendel_leisk]

third<div></div>

[mendel_leisk]

fourth<div></div>

[mendel_leisk]

Incidentally, Olivier, this raw conversion workflow also works very nicely with slide scans. With 0/0 wp and bp and no adjustment of gamma, results seem neutral and balanced.

[og]

Thank you Mendel for those tests.<p>

 

I have to say that I am a bit surprised that doing direct Gamma corrections on the Raw file does produce strange results. I will have to test this as well in my own workflow. Thanks for the input.<p>

 

From a theory point of view, I think that the gamma correction of one film is constant, and independant from Exposure (cf exposure/density curves of film). Even if brightness is very different on different photos, you should be able to apply the same Gamma correction - although film-processing may have an effect...<p>

 

I have to do some more testing to finalize my workflow for Color Negatives and Black-White Negatives, and to address those issues. I will post here the results (by the end of the year, hopefully - not so far : )<p>

 

Now, on your pictures:<br>

With your new Raw workflow, I think you have avoided the clipping of highlights in the upper part of the hand (which also emphasized grain on the Vuescan's one). I also prefer the tonal gradations, probably because of a slightly higher contrast on the new Raw version.<p>

 

So your new workflow is better, as far as I can see.<p>

 

About Vuescan: wouldn't you avoid this clipping of highlights by setting WP and BP clipping at 0.00% ?<br>

I know that 0.02% is probably not going to improve results a lot with Vuescan, but some people may argue [i do not... : ) ]<p>

 

Thank you,<p>

 

Olivier

[mendel_leisk]

I've run this raw conversion workflow on about half of one roll, and the gamma adjustment, to match Vuescan, was different in each case, whatever that proves. In the crop tab, I have left buffer on default 15%. This effects exposure calculation, dependent on how much the periphery varies from the central region. I'm not thrilled with this variety, it prevents automated workflow, but for now can't see a way round it.

 

I agree, the clipping is what is aggravating the grain. With Vuescan, highlight blowout is partly dependent on what you set in the colour tab, but there are more factors. Vuescan is at work behind the scenes when you scan as b/w negative, tweaking the profile.

 

I have images with very dense highlights. With the contrast index left at D76ci:.55, I will get blowout, lots, (you can turn on the colour indicators, to show where), regardless of white point setting. By raising to D76ci:.70, or even D76ci:.80, and bumping brightness, I can largely reduce or eliminate the blowout, BUT, the image is very, very flat. Similarly with black point, there is invariably an uptick at that end of the histogram, regardless of settings, unless I start playing with Vuescan's "advanced workflow". In essence, when you set white or black point, you're just saying, IF there isn't any blowout, introduce some.

 

Scanning with Vuescan with media type "image" largely eliminates clipped ends but I have two problems with this approach:

 

1. Poor feedback. I find it impossible to adjust brightness and contrast on a negative image.

 

2. The inverted output seems to have very washed out highlights (once I invert the image), regardless of what I set.

 

Regarding Vuescan scanning with media type set to "b/w negative", I strongly suspect some sort of reverse s-curve is being applied, and I'm getting tired of fighting it. It's too bad Ed doesn't provide some sort of "negative image" setting. By just converting the raw, I can get histogram that retains the flat toes and shoulders, NO blowout, AND better mid tone contrast.

 

I guess it goes without say but, to really see the difference in the samples I posted, you need to download them and view with some program that allows you to toggle between. The crops are positioned exactly the same, the tone difference is very apparent when toggling.

 

Oh well, all the best, it is Christmas eve, and I should drag myself away from the keyboard.