Mark Segals article: Scanning Colour Negatives: Raw of Not?

[Christoph C. Feldhaim]

Article Link: http://www.luminous-landscape.com/techniques/scanning_colour_negatives_raw_or_not.shtml

Thanks a lot, Mark, for taking the effort and writing up the quirky way of scanning color negatives and getting rid of the orange mask. I myself have a sort of hate-love relation to scanning for various reasons I won't disclose here for time reasons and maybe its just not important. However one question comes to my mind, since I do negative  neutralization a bit different:

I use the levels tool in PS to correct colors. I scan the negative as negative in 48 bit color (64 bit with infrared doesn't work for me, since its the Windows/LS-9000 issue ...). I also scan a bit of orange mask together with the negative, blur a portion of it, invert the image and then used this blurred part of the orange mask (the small stripe from between the negatives) as a reference point for the levels tool grey point to neutralize it. I just thought that is the quickest and easiest way, but maybe you have to say something about that.

[Bart_van_der_Wolf]

Article Link: http://www.luminous-landscape.com/techniques/scanning_colour_negatives_raw_or_not.shtml

Thanks a lot, Mark, for taking the effort and writing up the quirky way of scanning color negatives and getting rid of the orange mask.

Hi Christoph,

Unfortunately, there seems to be a fundamental issue with the base scan as apparently produced by Silverfast. On a capable scanner, and with a software that uses the possibilities of different exposure times per channel, there should not be an orange mask in the scanned data to begin with. Underexposing (!) the scan in the blue and green channels will increase scanner noise, and the subsequent 'repair' jobs will again increase noise and possibly contaminate colors due to sub-par removal of the mask in linear gamma space. The best quality can only be obtained if the mask is removed by different exposure times of the different channels.

Of course, when a scanner doesn't offer the capabilities of different exposures per channel, then one needs to make due with that.

Cheers,
Bart

[JohnBrew]

Mark, all that work and no 'red-eye' fix?

[Christoph C. Feldhaim]

Hi Christoph,

Unfortunately, there seems to be a fundamental issue with the base scan as apparently produced by Silverfast. On a capable scanner, and with a software that uses the possibilities of different exposure times per channel, there should not be an orange mask in the scanned data to begin with. Underexposing (!) the scan in the blue and green channels will increase scanner noise, and the subsequent 'repair' jobs will again increase noise and possibly contaminate colors due to sub-par removal of the mask in linear gamma space. The best quality can only be obtained if the mask is removed by different exposure times of the different channels.

Of course, when a scanner doesn't offer the capabilities of different exposures per channel, then one needs to make due with that.

Cheers,
Bart

I'll definitely try that out. My Nikon LS-9000 and Silverfast allow the regulation of each LED (R,G,B) luminosity.

[Kerry L]

Thanks Mark,
This is a great follow-up to the session at Vistek. I hope that there will be a second article on scanning MF transparencies including fluid mounting.

Questions:  In MF film, especially the positive emulsions, some types have inherently more contrast.  Should DR be controlled in the initial scanning or adjusted in post processing? As a follow-up, will the higher contrast films be noisier in the “shadow” areas?

[Mark D Segal]

Article Link: http://www.luminous-landscape.com/techniques/scanning_colour_negatives_raw_or_not.shtml

Thanks a lot, Mark, for taking the effort and writing up the quirky way of scanning color negatives and getting rid of the orange mask. I myself have a sort of hate-love relation to scanning for various reasons I won't disclose here for time reasons and maybe its just not important. However one question comes to my mind, since I do negative  neutralization a bit different:

I use the levels tool in PS to correct colors. I scan the negative as negative in 48 bit color (64 bit with infrared doesn't work for me, since its the Windows/LS-9000 issue ...). I also scan a bit of orange mask together with the negative, blur a portion of it, invert the image and then used this blurred part of the orange mask (the small stripe from between the negatives) as a reference point for the levels tool grey point to neutralize it. I just thought that is the quickest and easiest way, but maybe you have to say something about that.

Hi Christoph,

Yes, different things work differently with different images. That's part of the fun and the exasperation scanning negatives. What you are doing can obviously work. Nonetheless, my research indicates that within the family of the "shaken-down approach" (referring to that section of my article) into which your procedure falls, I have found that using a Curve in color mode, working with the black point setting, followed by another Curve in Normal or Luminosity mode for adjusting tonality, is the most systematically reliable way of getting the hue balanced without mucking-up the luminosity. When you use levels in RGB mode, it is a mixture of hue and luminosity and you have less control over each aspect individually. This can matter, again depending on the image.

[Mark D Segal]

Mark, all that work and no 'red-eye' fix?

Ahah - I left it there on purpose to see whether it would be noticed, and sure enough it was - a mark of attentive readership :-). OK, the real answer is: yes of course, I know it's there and very obviously so. And I did leave it there on purpose: (a) not a necessary adjustment in the context of the purpose of the article, (b) I actually wanted to use it, to see how the different approaches affect it - a little bit of good red data to throw into the evaluation mix.

[Mark D Segal]

Thanks Mark,
This is a great follow-up to the session at Vistek. I hope that there will be a second article on scanning MF transparencies including fluid mounting.

Questions:  In MF film, especially the positive emulsions, some types have inherently more contrast.  Should DR be controlled in the initial scanning or adjusted in post processing? As a follow-up, will the higher contrast films be noisier in the “shadow” areas?

Hi Kerry, transparencies don't pose the same challenges as negatives, because for those we have scanner profiles that are supposed to do a decent job of rendering the colours properly. But you have fingered a real problem with transparencies - the lower end of the luminosity range can indeed be very dense. And more density would normally translate into more noise. I've demonstrated in a previous article on this site (http://www.luminous-landscape.com/tutorials/scanning-old-new.shtml) that provided the blacks are not clipped in the scanning process, one can rescue detail either at the scan stage or the post scan stage, particularly in Lightroom where the combination of Fill and Blacks can be very effective.

That said, to minimize the risk of not having all the information one can reveal, and to reduce the heavy lifting in post-scan processing, I think it's sensible to take advantage of the luminosity adjustments available at the scan stage as a first pass at the issue. I normally use an industrial-strength noise and grain reducer at the post-scan stage (my favorite these days is Topaz Denoise) for handling noise and grain, and then use PK Sharpener Pro (often the 6*6 Positive sharpener) for sharpening the cleaned-up image. The combination of these two used in sequence gives one a lot of flexibility and quality editing.

I'm not likely to write a companion piece on transparencies in the near future, but later on, perhaps.

[Mark D Segal]

Hi Christoph,

Unfortunately, there seems to be a fundamental issue with the base scan as apparently produced by Silverfast. On a capable scanner, and with a software that uses the possibilities of different exposure times per channel, there should not be an orange mask in the scanned data to begin with. Underexposing (!) the scan in the blue and green channels will increase scanner noise, and the subsequent 'repair' jobs will again increase noise and possibly contaminate colors due to sub-par removal of the mask in linear gamma space. The best quality can only be obtained if the mask is removed by different exposure times of the different channels.

Of course, when a scanner doesn't offer the capabilities of different exposures per channel, then one needs to make due with that.

Cheers,
Bart

Hi Bart, this is interesting. Are you saying there is software that adjusts-out the influence of the orange mask of a colour negative at the scan stage automatically by using the lamp exposure controls of the individual RGB channels (where such capability in the scanner itself allows)? Or is this something the user needs to do manually with each scan provided the software allows it. "Lamp Brightness" in SilverFast, by the way, does allow this level of control, but it's comparatively awkward to use and I'm not convinced it yields better results than the other tools for this purpose. You need to find the right combination of RGB adjustments in the lamp control dialog to exactly neutralize the cast and this isn't a cake-walk. Do you have in mind another scan software you've used for this purpose which does this process more conveniently and yields better results which you've observed?

[Bart_van_der_Wolf]

Hi Bart, this is interesting. Are you saying there is software that adjusts-out the influence of the orange mask of a colour negative at the scan stage automatically by using the lamp exposure controls of the individual RGB channels (where such capability in the scanner itself allows)? Or is this something the user needs to do manually with each scan provided the software allows it. "Lamp Brightness" in SilverFast, by the way, does allow this level of control, but it's comparatively awkward to use and I'm not convinced it yields better results than the other tools for this purpose. You need to find the right combination of RGB adjustments in the lamp control dialog to exactly neutralize the cast and this isn't a cake-walk. Do you have in mind another scan software you've used for this purpose which does this process more conveniently and yields better results which you've observed?

Hi Mark,

I've owned a number of scanners through the years, and still have a few in operation. Some of those came bundled with Silverfast software, but I found the per scanner licencing and upgrade prices becoming a burdon. I therefore switched to VueScan Pro, and have been using it since many years on all scanners (even ones no longer supported by the manufaturer on new Operating Systems) without the need for re-learning (same user interface for all scanners, save differences in functionality). VueScan intimately operates at the scanner hardware level (after reverse engineering the command structure) and also produces Raw scans. In this case it is really Raw data, directly from the scanner interface, as provided by the scanner hard/firmware. VueScan therefore promotes the saving of such Raw data as a "Digital Negative", thus reducing the need for physical handling of the film material after the initial scan. That allows to benefit from later improvements in the software handling of the Raw data.

Part of the "advanced" workflow for masked negative films consists of determining the mask color of unexposed film (between images of from the leader) once per film (is constant per film, but may vary between batches), and then neutralizing it by adjusting the per channel exposure time (for scanners that allow that, e.g. Nikon Coolscan series). That results in boosting the blue and green exposure times, and as always this reduces the shot noise by maximizing the signal level for the most transparent film areas without clipping (result is maximum S/N ratio, also for the denser parts of the film). In addition it thus neutralizes the mask in linear gamma space (as it is digitized to Raw). What remains is Raw data with absolutely neutral shadow data. Any sensitivity difference between the color channels (non-parallel characteristic curves) will usually be rather linear in linear gamma space, so now it's easy to make the highlight (dense film area) data neutral with a simple levels kind of adjustment in linear gamma.

After these color calibrations, it is relatively trivial to invert and gamma adjust, plus adjust for the film tonecurve. The colors will remain neutral, and saturation can be adjusted at will. VueScan uses a database of film curves to do the tonemapping, but because of the predictable neutral state of the Raw data (can be saved as a TIFF) it also makes it easy to do in Photoshop if one creates one's own film curve.

I suppose Silverfast also offers such a facility, but I have no idea how easy it is to use. With VueScan it's quite simple.

Cheers,
Bart

[Mark D Segal]

Hi Mark,

I've owned a number of scanners through the years, and still have a few in operation. Some of those came bundled with Silverfast software, but I found the per scanner licencing and upgrade prices becoming a burdon. I therefore switched to VueScan Pro, and have been using it since many years on all scanners (even ones no longer supported by the manufaturer on new Operating Systems) without the need for re-learning (same user interface for all scanners, save differences in functionality). VueScan intimately operates at the scanner hardware level (after reverse engineering the command structure) and also produces Raw scans. In this case it is really Raw data, directly from the scanner interface, as provided by the scanner hard/firmware. VueScan therefore promotes the saving of such Raw data as a "Digital Negative", thus reducing the need for physical handling of the film material after the initial scan. That allows to benefit from later improvements in the software handling of the Raw data.

Part of the "advanced" workflow for masked negative films consists of determining the mask color of unexposed film (between images of from the leader) once per film (is constant per film, but may vary between batches), and then neutralizing it by adjusting the per channel exposure time (for scanners that allow that, e.g. Nikon Coolscan series). That results in boosting the blue and green exposure times, and as always this reduces the shot noise by maximizing the signal level for the most transparent film areas without clipping (result is maximum S/N ratio, also for the denser parts of the film). In addition it thus neutralizes the mask in linear gamma space (as it is digitized to Raw). What remains is Raw data with absolutely neutral shadow data. Any sensitivity difference between the color channels (non-parallel characteristic curves) will usually be rather linear in linear gamma space, so now it's easy to make the highlight (dense film area) data neutral with a simple levels kind of adjustment in linear gamma.

After these color calibrations, it is relatively trivial to invert and gamma adjust, plus adjust for the film tonecurve. The colors will remain neutral, and saturation can be adjusted at will. VueScan uses a database of film curves to do the tonemapping, but because of the predictable neutral state of the Raw data (can be saved as a TIFF) it also makes it easy to do in Photoshop if one creates one's own film curve.

I suppose Silverfast also offers such a facility, but I have no idea how easy it is to use. With VueScan it's quite simple.

Cheers,
Bart

Hi Bart,

SilverFast has all those same capabilities, but the use of the lamp exposure control is a bit awkward. I also have Vuescan Pro, so in the coming week I shall try to examine your procedure with my Nikon 5000, where it workd. I do know that fiddling with RGB controls to neutralize a cast is an iterative process, so one question really is how easily this can be achieved, which partly depends on the software interface.

Of course there is no real raw data in scanning as there is from digital cameras, except in the sense that it is unadjusted, and the film itself is non-linear from the get-go, but I see what you are getting at from the scan stage onward. I have found linear scans hard to work with in post-scan image editors, but I can also relate to the advantage you mention. More anon.

Mark

[David Mantripp]

Mark, first I'd like to thank you & Michael for keeping the topic of scanning alive and well, and for a very interesting article. It is interesting that in the end - arguably - the best approach is to use the tools Silverfast provides. I'm always coming back to this myself, discovering and rediscovering the strengths of it's adjustment tools, as compared with Photoshop. Sometimes they seem a bit eccentric, but I guess that's simply because they're different.  One very pedantic point - since you can apply NegaFix from SF HDR, maybe the best approach is to make an "archive" scan in 48bit/64bit HDR)i) anyway ? Then you've always got a fallback.

David

[Mark D Segal]

Mark, first I'd like to thank you & Michael for keeping the topic of scanning alive and well, and for a very interesting article. It is interesting that in the end - arguably - the best approach is to use the tools Silverfast provides. I'm always coming back to this myself, discovering and rediscovering the strengths of it's adjustment tools, as compared with Photoshop. Sometimes they seem a bit eccentric, but I guess that's simply because they're different.  One very pedantic point - since you can apply NegaFix from SF HDR, maybe the best approach is to make an "archive" scan in 48bit/64bit HDR)i) anyway ? Then you've always got a fallback.

David

David, thanks; yes indeed that makes complete sense.

Mark