By fabricator4/Chris (3,025) 
on June 19, 2005 2:56:46 AM CDT
Introduction
Scanning techniques
Masking as a method to alter scan characteristics
Normal unmasked scan
Blue masked scan
Orange masked scan
Further implications and thoughts
Conclusions and a few last thoughts
Introduction
One of the common complaints from B/W film users is that the scanning technology is not up to the task of dealing with a well developed negative with a wide tonal range. This obvious shortcoming concerns me somewhat because I have been using an Epson 2480 scanner for basic sensitometry and testing. Another technique that suggests itself is doing two scans, one for shadows and one for highlights, but this seems unnecessarily messy, and according to those that have tried it not entirely successful. After participating in this forum post I was prompted to do some testing along some lines suggested, namely using a colour filter mask to modify the scan characteristics. Where the testing and subsequent thought led me was a startling revelation, finally giving me an answer to the questions raised but by a completely different means.
Scanning techniques
Scanning a B/W negative is anything but straight forward so it's necessary to outline my methods here since they may differ from someone else's. Part of the problem is that while most scanning software purports to have the ability to scan B/W negatives it rarely does a credible job on automatic. The Epson Scan software (V 2.40) seems to simply use the same algorithms as it does for colour negatives with only lip service paid to the higher dynamic range, and I suspect from comments I've heard that other software does the same. The results are invariable a flat image, that when you attempt to adjust the contrast loses any attractive tonal range you pre-visualised, especially in the shadows.
My own technique is to adjust the white and black to maximum and minimum respecitively and then set the scan histogram to the appropriate white and black points. This in my mind sets the exact points at which I want white and black on the final image. One of the problems with this is that the Epson software then bends the response curve, which I adjust by changing the gamma. This at least results in an image the approximates what I pre-visualised however for the most difficult negatives it's not in itself enough. This brings us to the current article, and all of the scans presented here are the result of this technique.
Masking as a method to alter scan characteristics
Normal unmasked scan
Here's the normal unmasked scan using the techniques described in the last section. This image came about during a camera test: The tree was backlit by the bright clouds so it seemed to be surrounded by a halo, and the light was quite inspiring. While I realised at the time I was asking a lot from my film I also knew I had the 12 stop dynamic range to attempt it if I exposed for the incident light falling around me, which I did. The negative does indeed have all the inspiring light and detail I was hoping for, however as you'll see the results here and in my folio are much less than inspiring.
Notice firstly the nice highlight detail in the clouds. It was really quite bright so I was pleased at the level of detail and tonal range. There's actually much more than this on the negative, see my first folio post of this image. In an effort to scan this detail however what has been sacrificed is all the other tonal information. The white fence is grey, the white tree is even darker, and the grass is a muddy grey. Even worse, the foliage at the end of the field had disappeared into a dark morass. One of the "features" of my method of scanning this negative is badly blocked up shadows: When I adjust the gamma to make the response curve flat it jams a lot of detail down there at the toe of the curve and blocks it up as a result.
Blue masked scan
It was suggested that an orange masked be used but I had a brainwave to try other colours as well. This is the result of putting a 80A colour correction filter over the light source. You'll have to excuse the vignette because it's a 55mm filter over a 6x6 negative.
I guess I presented this one first not just because it's the first one I tried, but also because it's an abject failure. The shadow detail is even worse, lost in the much higher noise level I'm sure. Despite this however I think the amount of detail in the mid to dark tones is still there despite the noise. My impressions from the pre-scan were that it was worth pushing on with the test because it looked like it did make an improvement before the noise messed it up.
Orange masked scan
I didn't have the orange film mask that I needed to do this test until I remembered a Cokin sunsoft filter I'd been given by a supplier as a freebie. The colour of this filter is extremely close to the colour mask of some C-41 films I've used:
While the results are not earth shattering it's interesting to note the greater detail in both highlight and shadows. Look at the bush line immediately behind the tree. This extra detail can make a big difference to what you can do with an image like this when editing. The higher noise level is an unfortunate side affect and relates to either the shortcomings of the entry level scanner or the technique, or quite possibly both. The filter is also a soft focus type and it may have imparted some of it's own qualities. This exercise would need to be redone with a clear orange film base taped over the light source to be sure. There's no denying the technique has possibilities however.
Further implications and thoughts
All of this was very interesting and shows some degree of promise, however it still didn't answer the question of why I can use the 2480 accurately enough to perform some sensitometry (see the characteristic curve plots and how-to article in my journals) and yet cannot scan an image with the large dynamic range previously recorded in my curves. The answer seems to be that while the scanner is able to see the complete tonal range, what it does with it during the scan is another matter. Like when much of the shadow detail gets jammed into the toe of the curve, it seems to apply a curve of it's own devising when scanning (and not just sensing) a complete negative. I could use the sensitometer function on this negative and see that the shadows were on about zone I and II, but what the scanner software was doing to this was something else. Differences of up to 10 luminance points were being compressed into 2. This was on a scan which was unadjusted to my knowlege. This correleates also to the response curve in the histogram windows. It shows and 'S' curve and tries to increase the 'S' as I adjust the white and black points (remember my scan procedure as outlined above).
While pondering this I started to wonder if the software used the same response curve for slides as it did for negatives (surely not!) and it also occurred to me that the dynamic range of slides is as great (or greater?) than we have with B/W negatives. Perhaps if the B/W scans fail because the software authors are mistakenly using the same response pattern for both colour and B/W negatives, perhaps scanning the B/W negative as a slide will give different results. Here it is.
Of course I had to change the negative image to a positive, and at first this didn't look too different but I had some idea what to look for because (and this was the first big revelation) when I adjusted the white and black points in the pre-scan the software maintained a straight response characteristic. Yes. Really. The holy grail I'd been looking for had been right under my nose all along. I was so excited in fact that I didn't notice if the gamma corrected itself (it didn't) and immediately did this scan. Look at the shadow detail in the bush line at the end of the field and you'll see the zone I and zone II information that's missing on most of the other scans. True, this is still not an earth shattering image scan, but it does get me another step in the right direction.
Conclusions and a few last thoughts
As I said, the scan is still not that wonderful. Part of the reason is that an 8 bit grey scale is still a bit of a squeeze when it comes to an image with so much dynamic range and detail. Trying to do it has forced me to squash middle grey (the grass) down a zone or two so the result is muddy to say the least. Perhaps in another five years we'll all be using 48 bit video systems, scanners and cameras as a matter of course, and I'll be able to look back and laugh. Just in case you're still in any doubt as to the affect of scanning a B/W negative as a slide I've done a couple of cropped views of just the tree line and a bit of the horizon. I think you'll agree the results speak for themselves:
:-)
Read 10,013 times
Reply
|
From bluetexasbonnie/Bonnie (8,302)
on June 19, 2005 9:19:18 AM CDT
GREAT! I'm pulling out some of my recent 'promising' B&W to rescan. Maybe their promise will be fulfilled.
I found it interesting (in an ugly sort of way) that using the orangish filter increased noise. That was exactly my complaint about color scan AND exactly what I saw when I sandwiched a bit of colored leader with my B&W.
It is so wonderful and kind that you are sharing all of your insight. I really appreciate it.
Reply
|
|
|
From fabricator4/Chris (3,025)
on June 19, 2005 10:25:50 AM CDT
Hi Bonnie. It occurred to me later that I should have also tried a slide scan with the addition of the orange filter. The slide scan on its own seems a lot cleaner although I haven't done any really definitive testing yet. It's 12:23 AM monday here and I have to get up in a little over five hours to go to work.
Reply
|
|
|
From zoneone/Stephen (2,210)
on June 19, 2005 4:28:46 PM CDT
OMFG this works. As soon as I read this I tried it out with my Epson 4870. This has got to be the best piece of advice for scanning I have seen in years.
Reply
|
|
|
From banister/John (834) 
on June 22, 2005 5:14:19 AM CDT
Thanks very much for the article. When I started reading your words in the "Further implications and thoughts" section, what came into my mind was a scanner saving output data in "scanner RAW" but then is sounds like you've already found a way of getting less processed output. I can remember having
some Scala that had details both in the shadows and in the highlights that seemed lost, but I was able to find each at the expense of the other by choosing different levels for middle grey in the prescan image before scanning. I don't know for sure that the scanner actually changed the amount of exposure time during the scans, but I had the impression that it did. I'm away from home right now, or I'd do some testing and check that again. Do you think you could
get what you want by making composites of scans optimized for different brightness levels?
Reply
|
|
|
From fabricator4/Chris (3,025)
on June 27, 2005 5:35:07 AM CDT
output data in "scanner RAW"
You can actually do this to a large extent by turning all automatic corrections off. What you get is a rather flat image that wastes a lot of the tonal range in an 8 bit file. Yes you can increase the contrast but you no longer have the full 256 shaded of grey. IOW if your RAW scan is so flat that all the tonal range is contained between 100 luminance points, after increasing the contras you will still only have 100 shades of grey. We can scan at 16 bits easily but there no software that will edit it. If there was we might still be able to have 256 shades in an 8 bit file and full controll over where those shades come from.
With the current level of technology I'm convinced the secret to doing this well is curves and histogram adjustments based on the pre-scan data. I'm equally convinced that the software authors don't have the faintest ideas what a B/W negative actually is.
Do you think you could get what you want by making composites of scans optimized for different brightness levels?
I'm sure I could, given enough time and I must admit the idea is very attractive. The problems seem to be in the re-combination of the images not giving the right mid tones. The Scan software does allow you store custom curves and that is one approach that I need to investigate further before I start testing multiple scans and trying to find useful way to recombine them.
Part of the problem is that there so many ways of doing this, and there doesn't seem to be a single "right" way that works perfectly for all B/W negatives. I've just done some portraits that scan quite nicely if I just follow my normal procedures for setting white and black points. It's the images where I set out to use the full dynamic range of the film (ie zone system exposures based on subject brightness) that give the problems.
Reply
|
|
|
From charlesheckel/Charles (2,297)
on June 26, 2005 5:18:53 PM CDT
Interesting. You didn't say anything about the software you use with your scanner. With my 2450 they packaged the Epson EOM software and another called Silverfast, which has quite a good reputation, but wasn't as intuitive as the scattergrams and sensitometric curves in PS.
My impression is that the neg-scanning software expects quite thin negs with a long toe, "read a newspaper through the highlights" kind of stuff, and the tranny-scanning software expects a large dynamic range and a much flatter curve for producing a duplicate image.
Reply
|
|
|
From fabricator4/Chris (3,025)
on June 27, 2005 6:00:55 AM CDT
Hi Charles, yes I'm using the Epson Scan softare, V2.50E. My 2480 is the entry level scanner and did not come with Silverfast. I believe that Silverfast does work a little better in some areas but I don't know whether B/W negatives is one of them.
Epson Scan has histograms and curves as well as the normal adjustments. The interface doesn't work quite as nicely as I would like. The "save current settings" for example doesn't seem to save and restore anything that I need it to like scan resolutions and other user parameters.
The save and restore custom curves does seem to work, but I haven't figured curves into my scan workflow yet. One of my main complaints as you've gathered is the way the software tries to interpret the B/W negative.
My impression is that the neg-scanning software expects quite thin negs with a long toe, "read a newspaper through the highlights" kind of stuff, and the tranny-scanning software expects a large dynamic range and a much flatter curve for producing a duplicate image.
That's pretty well right from what I can tell. The problem is my processing setup is based on the speed tests and dynamic range tests, rather than how thin the negative is. My zone one exposure is already on the straight part of the charactistic curve - the toe is quite distinct, and this is probably the case for most people following published processing times. Chromogenic (C-41) B/W film on the other hand would be a completely different animal, and may possibly scan the way the software authors intend.
Reply
|
|
|
From mp165/Mike (1,142)
on September 6, 2005 12:37:23 AM CDT
brilliant! i use a minolta dimage scan dual III. i've been having trouble getting detail in hilights on my B&W negs for a while. thin negs were great - anything else ended up looking extra grainy and blown out. now i can scan it as a colour negative for the best result, or as a slide for a close second. if you're ever in calgary i owe you a beer! ;)
- mike
Reply
|
|
|
From altan1/Altan (25)
on September 24, 2005 4:08:22 PM CDT
Hi Chris, I'm wondering, what software do you use for scanning? The software that comes with a scanner is typically not that good, specially if you have a flatbed scanner that "also" scans negatives. I'm not sure if I'm alowed to mention commercial products on this forum, but I would suggest anyone who is serious about scanning negatives or slides to look at "vuescan". I've been using it for years and have scanned thousands of negatives with it.
Reply
|
|
|
From fabricator4/Chris (3,025)
on December 23, 2005 8:58:55 PM CST
I've got vuescan, and while it's very good I often find I get better results with the Epson Scan software. Part of the problem with vuescan is that I get heavy scan lines that don't show up with Epson Scan. Obviously the vendor software is doing stuff that Vuescan doesn't know about to get the best out of the hardware.
Vuescan has some very nice features but I'm kind of sorry I bought it at this stage, becuase I can't use it for a lot of my stuff and it's really a waste of time. Maybe when I get a better scanner it will start paying for itself.
Reply
|
|
|
From liversb/Bill (38)
on December 28, 2005 6:28:21 AM CST
Hi Chris,
I too, have be using Vuescan for some years now and so far have been able to tweak it to produce better colour scans than the Epson software (3200) especially with the histograms. Wish there was this sort of forum for it, as I am sure that many others have done exactly what I have done (but probably only spend 50 hours getting there). Ed Hamrick still produces several new versions every month, so he is active on the development side. It may be worth mailing him and see if he is able to do anything?
Reply
|
|
|
From mitkase/Mit (0)
on March 15, 2006 1:11:35 PM CST
wow--thanks for the insightful tips! my scans (from a perfection 3170) certainly *look* better; however, and i wonder if anyone else has noticed this, epson scan 2.65A no longer correctly identifies the size of the negative when set to positive film. it cuts seemingly randomly across the image. a quick change to b/w negative and the cutting is perfect... so that out of two rows of film, each time i can only use one or two of the images... and moving the images seems to do nothing; this seems to be dependant on the composition and coloring of the image. strangeness, indeed. has anyone else run into this problem?
also. in light of this, i'm wondering if we should reconsider other aspects of the scan software. has anyone played around (again) with whether to scan in grayscale or color and convert later?
Reply
|
|
|
From fabricator4/Chris (3,025)
on May 17, 2006 1:31:56 PM CDT
You are likely to run into problems with framing if the negative is very light. I don't know why the neg scans finds the frame OK but slide scan does not. Perhaps it's not recognising the clear space between frames as the boundary since it's supposed to be reversed for a slide scan. The solution is to use normal preview instead of thumbnail. You can then define the frames yourself and zoom in to adjust the scan.
My preference is to scan 48 bit color even for B/W negs. Conversion to monochrome can be done later. I've found grain and noise is emphasised if you scan 8bit B/W and let the scanner/software do the conversion.
Reply
|
|
|
From mitkase/Mit (0)
on March 15, 2006 1:37:10 PM CST
wow--thanks for the insightful tips! my scans (from a perfection 3170) certainly *look* better; however, and i wonder if anyone else has noticed this, epson scan 2.65A no longer correctly identifies the size of the negative when set to positive film. it cuts seemingly randomly across the image. a quick change to b/w negative and the cutting is perfect... so that out of two rows of film, each time i can only use one or two of the images... and moving the images seems to do nothing; this seems to be dependant on the composition and coloring of the image. strangeness, indeed. has anyone else run into this problem?
also. in light of this, i'm wondering if we should reconsider other aspects of the scan software. has anyone played around (again) with whether to scan in grayscale or color and convert later?
Reply
|
|
|
From mitkase/Mit (0)
on March 15, 2006 2:11:06 PM CST
wow--thanks for the insightful tips! my scans (from a perfection 3170) certainly *look* better; however, and i wonder if anyone else has noticed this, epson scan 2.65A no longer correctly identifies the size of the negative when set to positive film. it cuts seemingly randomly across the image. a quick change to b/w negative and the cutting is perfect... so that out of two rows of film, each time i can only use one or two of the images... and moving the images seems to do nothing; this seems to be dependant on the composition and coloring of the image. strangeness, indeed. has anyone else run into this problem?
also. in light of this, i'm wondering if we should reconsider other aspects of the scan software. has anyone played around (again) with whether to scan in grayscale or color and convert later?
Reply
|
|
|
From ully2000/Richard (1)
on May 25, 2006 9:58:20 PM CDT
Try the Kodak C41 B&W. I have had pretty good luck with it. Before that I did the T400CN and it was as pretty good too. Have never had very good luck on my Epson 4490 with TriX.
Reply
|
Return to articles
|
