PDA

View Full Version : How to build a scanning curve?



Stephen Sample
2-Dec-2006, 17:36
In the excellent January/February 2006, View Camera Magazine article, A Simple Way to Test For Film Speed and Developing Time, Steve Simmons refers to a final step of building a scanning curve. This curve is then used to bring in your image as it proofed for you. Can anyone direct me to a web site or book which will instruct me on this? I think I understand what a film curve is, but am new to scanning and digital printing. Thanks

Chris_Brown
10-Dec-2006, 21:47
In the excellent January/February 2006, View Camera Magazine article, A Simple Way to Test For Film Speed and Developing Time, Steve Simmons refers to a final step of building a scanning curve. This curve is then used to bring in your image as it proofed for you. Can anyone direct me to a web site or book which will instruct me on this? I think I understand what a film curve is, but am new to scanning and digital printing. Thanks

I can't direct you to a web site with instructions, but I can help get you started.

The idea is to develop a set of input parameters in your scanning software that gets you acceptable results without a lot of fussing and guessing with that software.

To do this, set up a controlled scene with:

A MacBeth Color Chart, for the gray scale and color swatches
A clean, polished chrome object, for Dmin
A large area of black velvet,for Dmax
A person, for flesh tones

Shoot a bracket of this scene and process in your normal fashion. Make sure you color balance your film for your lights. Mount the film on your scanner and

Use the black velvet to set the shadow endpoint
Use the polished chrome to set the highlight endpoint
Use the gray scale to set the RGB curves for neutral gray throughout the density range
Use the color swatches to verify your RGB values (MacBeth color charts have very useful RGB info on the back of the chart)
Use the person to verify correct RGB skin tones and to visually evaluate the scan

After you've made a few test scans and have dialed in your settings, save those settings for that film type. Then, when it comes time to scan a batch of that film, you've got an excellent starting point.

I should point out that Don Hutcheson (http://www.hutchcolor.com/index.html), a revered scanning expert, is a proponent of scanning everything in 16-bit color using only one setting that accomodates your scanners total input gamut. His arguments for this are pretty good and worth a read.

I hope this helps.

Stephen Sample
11-Dec-2006, 21:28
Thanks Chris,

I plan to try your suggestions soon. Also, the Hutcheson web site had some wonderful scanning resources and articles. I appreciate your input. Steve

Leonard Evens
12-Dec-2006, 07:01
I can't direct you to a web site with instructions, but I can help get you started.

The idea is to develop a set of input parameters in your scanning software that gets you acceptable results without a lot of fussing and guessing with that software.


These suggestions are quite reasonable, but let me note that they assume you are using transparency film. For negative film, color or b/w, the second and third should be reversed.



To do this, set up a controlled scene with:

A MacBeth Color Chart, for the gray scale and color swatches
A clean, polished chrome object, for Dmin
A large area of black velvet,for Dmax
A person, for flesh tones

Shoot a bracket of this scene and process in your normal fashion. Make sure you color balance your film for your lights. Mount the film on your scanner and

Use the black velvet to set the shadow endpoint
Use the polished chrome to set the highlight endpoint
Use the gray scale to set the RGB curves for neutral gray throughout the density range
Use the color swatches to verify your RGB values (MacBeth color charts have very useful RGB info on the back of the chart)
Use the person to verify correct RGB skin tones and to visually evaluate the scan

After you've made a few test scans and have dialed in your settings, save those settings for that film type. Then, when it comes time to scan a batch of that film, you've got an excellent starting point.


Let me also add that a typical gray scale won't usually include a sufficient range of values. Instead, I would recommend varying lighting over the scene and spot metering various uniformly lit areas in the scene. For negative film, you should aim for as many EV values as possible over a an 8 to 10 stop (or higher) range. For transparency film, a narrower range of EV values might be acceptable.

Leonard Evens
12-Dec-2006, 07:12
Let me raise another point, which I've been trying to emphasize, without apprent success.

When you scan, the software typically will allow you to read RGB values in the range 0 to 255. By modifying various scanning paremeters, you can change these values, and they are what you need to set correctly so the result will look the way you want it to, either on your monitor or in a print. (Printing also will require calibration.) Unfortunately, the scanning software often seems to have a will of its own in setting these RGB values. Very similar scenes may yield different such values even when you think everything has been standardized. For this reason, it is helpful to be able to read the raw densities above base. This is closer to what the scanner reads from the film and which the scanning software uses to determine the RGB values by a somewhat complicated algorithm. Some scanning software, such as Vuescan, allows you to read the densities above base. I find it easier to get a consistent idea of what is on the film using these.

Ted Harris
12-Dec-2006, 07:25
Stephen, the suggetions so far are good. Especially don't ignore doing all your scanning in RGB and that this method is a starting point. Once you set up a workflow of this sort you still need to evaluate each individual negative/transparency and make the best setting sfor that particular piece of film, just as you would do if you were enlarging the film. For many properly exposed peices of film this will simply mean setting the white and black point for that particular image and you will often find that your software (especially if you are using Silverfast Ai) has already chosen the best points but you do need to do this for every scan if you want optimum results.

Bruce Watson
12-Dec-2006, 07:26
In the excellent January/February 2006, View Camera Magazine article, A Simple Way to Test For Film Speed and Developing Time, Steve Simmons refers to a final step of building a scanning curve. This curve is then used to bring in your image as it proofed for you. Can anyone direct me to a web site or book which will instruct me on this? I think I understand what a film curve is, but am new to scanning and digital printing. Thanks

I understand doing this for scanning transparencies. For negatives, either color or B&W, it doesn't make much sense to me.

The reason is that a properly exposed tranny usually has a density range that is both large and relatively consistent. It has to if it is to perform its primary task which is image projection. It has to have a black that is black and a white that is white. Not relatively black or white, but actually black or white. Since the density range is relatively predictable, it makes some sense that one could find a scanner setup that will work relatively well with most every tranny.

Not so with negatives. The density range of negatives varies widely. It's not possible to find a single scanner setup that will work relatively well with most every negative, particularly B&W negatives.

The best workflow I've been able to come up with for drum scanning my negatives is to set the black and white points for each channel (R, G, or B) individually, setting the log amp circuits to give their full range of responses to these limited density ranges. This results in a scan file that is very close to the proper contrast for each color, and minimizes color correction work in photoshop for me. This works particularly well for my B&W scans.

Profiled scanners typically result in negative scans that are both flat and sparse. By sparse I mean that the histograms show that the scan data doesn't fill the digital range well. This means one has to chop off the empty ends of the histograms using the photoshop levels tool for example. This in turn raises the risks of posterization as photoshop has to spread this sparse data over a larger digital range.

I'm afraid I'm not being very articulate this morning. Sigh... What I'm saying is, if you are scanning trannies go ahead and make a profile for the scanner if you want. If you are scanning negatives, I think you'll be wasting your time. But you can try it and come to your own conclusions.

C. D. Keth
12-Dec-2006, 10:58
I beg to differ, Bruce. Unless one is doing something seriously wrong, most of one person's negatives of a given filmstock, processed the same way, and exposed the same way will tend to have very close d-min, d-max and gamma. Making a scanning curve for this situation will get you 95% to what you need, with the last 5% being for variation in subject, mostly. I think it's worth doing for the time it will save you making the same corrections over and over.

Chris_Brown
12-Dec-2006, 10:58
Let me also add that a typical gray scale won't usually include a sufficient range of values. Instead, I would recommend varying lighting over the scene and spot metering various uniformly lit areas in the scene. For negative film, you should aim for as many EV values as possible over a an 8 to 10 stop (or higher) range.

This assumes that the material being lit (e.g., white seamless) has no color bias and the lights used have been balanced to the film (i.e., targeted to a specific color temperature and green/magenta balance).

I prefer the MacBeth chart because the swatches are neutral and without color bias. If more than five data points are needed on a gray curve, I think it'd be a waste of time - adjustments will have to be made for the actual photos to be scanned (not the test image).


For negatives, either color or B&W, it doesn't make much sense to me. (...) The density range of negatives varies widely. It's not possible to find a single scanner setup that will work relatively well with most every negative, particularly B&W negatives.

So, you're saying you begin your neg scanning sessions with no presets in the scanning software? It seems you're re-inventing the wheel each time.

If the neg density changes in negs due to intentional variations in processing, then a test neg could be made for each push or pull scenario. That's a lot of work for the extra mileage, unless you push/pull frequently.

C. D. Keth
12-Dec-2006, 11:04
If the neg density changes in negs due to intentional variations in processing, then a test neg could be made for each push or pull scenario. That's a lot of work for the extra mileage, unless you push/pull frequently.

Exactly what I think. If you have THAT much variation in your processed negatives, there are more serious problems than your scanner curves.

Bruce Watson
12-Dec-2006, 12:18
So, you're saying you begin your neg scanning sessions with no presets in the scanning software? It seems you're re-inventing the wheel each time.

You could say that. I say that I'm making an exact fit to my negative each time, which is what I think a good scanner operator should be doing.

If you want a series of B&W negatives to all exhibit a consistent density range, and the SBR varies widely as it often does for landscape work for example, you have to do the work somewhere. To get a given density range you do the work in developing (aka Zone System style expansion and/or contraction). I find it's simpler, easier, and gives me better results to do the work in setting the black and white points for each negative. With my scanner software, this takes less than a minute. But if you want to call it "re-inventing the wheel" I guess you can.

What this is, is a fairly large religious debate. There are people on both sides. Some say to profile scanners, some say that profiling scanners is a waste of time. I think that profiling scanners for trannies might be interesting, but that negatives vary too much to make profiles very useful.

In other words, I don't think "one size fits all" is the optimum way to get the most from a scanner when scanning negatives. But if it works for anyone else I say go for it. Whatever floats your boat.

Bruce Watson
12-Dec-2006, 12:27
I beg to differ, Bruce. Unless one is doing something seriously wrong, most of one person's negatives of a given filmstock, processed the same way, and exposed the same way will tend to have very close d-min, d-max and gamma. Making a scanning curve for this situation will get you 95% to what you need, with the last 5% being for variation in subject, mostly. I think it's worth doing for the time it will save you making the same corrections over and over.

Permission to differ granted. This isn't a dictatorship!

That said I have hundreds of well processed 5x4 Tri-X negatives that scan beautifully but that have a wide variance of density range. Then again, I quit doing any expansion and contraction development years ago. With scanning, it just isn't needed. What works is the extremely simple system of "expose for the shadows and let the highlights fall where they may."

I've tested until blood ran out my ears (well, maybe not quite that far). I understand how it works and why it works, but only because I've done the testing and proven it to myself.

Kirk Gittings
12-Dec-2006, 12:48
What works is the extremely simple system of "expose for the shadows and let the highlights fall where they may.

I did not see what the poster was scanning with but I cannot disgree more with this statement particularly with consumer flatbeds.. We have discussed this earlier so I won't repeat myself, but this lack of developement control of the highlights in my experience leads to real problems particularly with the consumer flatbeads, a bit less so with Imacons and even less so with drum scans. I have negatives just like this that nobody can scan properly because the highlight detail is so bunched up there is no way to separate it.

Chris_Brown
12-Dec-2006, 14:47
If you want a series of B&W negatives to all exhibit a consistent density range, and the SBR varies widely as it often does for landscape work for example, you have to do the work somewhere. To get a given density range you do the work in developing (aka Zone System style expansion and/or contraction). I find it's simpler, easier, and gives me better results to do the work in setting the black and white points for each negative. With my scanner software, this takes less than a minute. But if you want to call it "re-inventing the wheel" I guess you can.

Ah. I think you misunderstand me. I'm simply using my test film to give me starting input values, not final values.

Example:
Acquire a density reading (from the scanner) of film base + fog from a normally processed and exposed test sheet of B&W film. Save this reading in the scanning software. Then, for example, when an image comes up where FB+F is not close to the test reading, use that knowledge to set endpoints and midpoints accordingly.


What this is, is a fairly large religious debate. There are people on both sides. Some say to profile scanners, some say that profiling scanners is a waste of time. I think that profiling scanners for trannies might be interesting, but that negatives vary too much to make profiles very useful.

I think profiling a scanner for transparency work is worth the time and effort. My transparency scans from both an Imacon and Howtek drum scanner improved markedly after I began using the profiles correctly.

"Experts" say that profiling is a waste of time with color negatives, because the orange mask compresses blues so much. Okay, but I've not read how these experts approach the rest of the color workflow (not yet, anyway). Does the use of scanned negatives eliminate the need for a color-managed workflow? I don't think so, especially when an image is to be printed on any modern, ICC savvy inkjet printer. Color management helps attain accurate color and tonal reproduction, whether the image comes from negs or chromes.

So the question becomes: Where does one begin the color-managed workflow if the scanner profile is not used?

If I take an untagged image and assign an Adobe '98 profile to it (for example), I am saying that the colorspace of film negative+scan=Adobe '98. This is a path to poor color/tonal reproduction. Somehow, the colorspace of the neg+scan should be found to provide an accurate entry point into a color-managed workflow.

And this is where religion comes in. There's plenty of prayers while the clock spins, as a 2.5 Gigabyte file opens in Photoshop. ;)

Kirk Gittings
12-Dec-2006, 15:04
With color negatives and Epson scanners with Silverfast, Michael Mutmansky and I have had reasonable success shooting a Macbeth chart, and scanning it in using one of SF canned film profiles that best approximates the Macbeth, tweaking the settings and saving that as a modified film profile. For myself I have found the need to create a series of these modified profiles depending on the lighting in the original scene I.E strobe lit interior, strobe fill interior with incandescent, strobe fill interior with flourescent, ambient light flourescent interior, exterior daylight, exterior twilight etc.etc. Though a simple daylight one really handles most variations adequately.

Bruce Watson
12-Dec-2006, 15:55
I have negatives just like this that nobody can scan properly because the highlight detail is so bunched up there is no way to separate it.

It doesn't sound to me like your problem negatives compared to mine are "just like this." It sounds like your problem negatives have a fair amount of image information off in the film's shoulder region which mine do not. If so I'll agree that they won't scan well. But then, nothing is going to salvage such negatives.

The underlying assumption I'm making is that one is using modern films and processing and hitting the exposure well so that the image runs from just above the film's toe up through the linear portion of the response curve. That is, "expose for the shadows" means the same thing in this discussion as it has meant for decades to Zone System practitioners.

I'm not talking about a slap-dash approach here. In order to expose the film correctly you still have to do your EI tests and your normal development tests (which for me gives me a Zone VIII density of around 1.0). Just like you would for the Zone System. Nothing I know of is going to let you off the hook of making a good exposure and having control of your process.

If you do that, you'll be fine. Modern B&W films seem to have rather long linear responses. "Expose for the shadows and let the highlights fall where they may" works fine for me with 5x4 Tri-X out to around 12 stops of SBR (a white magnolia blossom in full sun, midday in June). A more modern T-grain film would work just as well if not better. Negatives treated this way scan well on my old Epson 2450 consumer flatbed with VueScan. And they drum scan just beautifully.

Henry Ambrose
12-Dec-2006, 16:23
It seems this thread has wandered a good bit from the original question. If I remember correctly, the article that Stephen Sample referenced was about B&W negatives. So I assume that he is asking about B&W negative scanning. What I write about in this post is B&W negative scanning.

For work done in a controlled environment having a saved set of adjustments that is keyed to your film, processing, subject and lighting methods is a nice idea to make your scanning process faster. You get to start close to the best settings without having to re-set them each time you scan a different negative. I think what Steve Simmons meant in his article is literally a "curve" that you'd save. It depends on your software exactly how this is done. In Epson's PS plugin its in "Tone Correction" and you simply make adjustments in the curve window that will be shaped like an "s". (remember those film and paper curves in the darkroom?) When you get the preview to look like what you want, save that with a name that makes sense to you. Maybe "HP5 open shade" or "TMax100 studio" - something like that. Next time you scan HP5 shot in open shade you can select that tone curve from the pulldown menu. You'll be in the neighborhood of the right look and can fine tune from there.

As Bruce wrote, you're gonna have to set black and white points each time. You'll also have to be sure that the negative you expose and process contains densities that are within the capture range of -your- scanner. Anything outside of that range is lost. Kirk made note of this and it is vital that you follow this advice.

Now about color profiles - for B&W scanning there aren't any. Color work uses profiles to accurately translate color from device to device. "Coke can red" as seen by your scanner looks like "coke can red" on your monitor looks like "coke can red" on your print. Its that simple in concept. B&W does not use color profiles because there is no color to manage.

Chris_Brown
12-Dec-2006, 16:23
For myself I have found the need to create a series of these modified profiles depending on the lighting in the original scene I.E strobe lit interior, strobe fill interior with incandescent, strobe fill interior with flourescent, ambient light flourescent interior, exterior daylight, exterior twilight etc.etc. Though a simple daylight one really handles most variations adequately.

Just to clarify, when I refer to a "profile" I mean an ICC Color Profile generated from scanning an IT8 or HCT target transparency and processed through ICC Profile-generating software such as BasICColor Scan (http://www.colorsolutions.de/english/index_E.htm) or Gretag-MacBeth ProfileMaker Photostudio (http://usa.gretagmacbethstore.com/index.cfm/act/Catalog.cfm/catalogid/1982/Subcategory/ProfileMaker%20Solutions/category/ProfileMaker/browse/null/MenuGroup/_Menu%20USA%20New/desc/PM5%20Photostudio%20Pro%20%28software%20only%29.htm).

I'm not referring to the saved input settings (endpoint, curve points, USM values, etc.) of scanner software.

Chris_Brown
12-Dec-2006, 16:28
Now about color profiles - for B&W scanning there aren't any.

What if you scan B&W negs in RGB mode?

Henry Ambrose
12-Dec-2006, 16:30
I am sure that Bruce & Kirk both know what they're about. But maybe it's easier stated in this most basic way: "you must get the scene information onto your film in a way that your scanner can see it".

Henry Ambrose
12-Dec-2006, 16:33
What if you scan B&W negs in RGB mode?

OK, go do this and come back and explain how it worked for you.
; >)

Kirk Gittings
12-Dec-2006, 16:36
I understand that, but those will not work with color negs right? At least my conversations with the engineers at Silverfast indicated that they wouldn't otherwise they would be offering them (there is a huge demand for them, as they do IT8 transparency targets and processing software in SF AI Studio and no other sw/scanner manudfacturer offers them either as far as I know). The SF engineers would not tell me however, how they did generate their color negative film profiles. Hence tweaking the existing profiles is all we can do.....In SF AI Studio there is a whole separate set of detailed information and tools in the Negafix Expert Dialoge for interpreting and modifying the scans and saving modified profiles.

Henry Ambrose
12-Dec-2006, 17:24
Its possible to make an internal profile of sorts for color negative but it will only apply precisely to the particular piece of film and particular conditions of exposure and processing. So if you try to use it for a bit different exposure or the lab just changed the chemistry its not accurate. There's a whole 'nother step involved in converting the negative color image into positive thats outside of normal ICC profile building. But at least they do give us internal film profiles that get us started down the right path. Internal meaning that its for use by the scanning software to convert the negative to some positive colors and not by the operating system color management.

You have a positive thing in your hand that is a particular color to refer to. The light blue (or whatever color) in your color transparency is the reference. That actual color is described mathmatically when it is digitized. Numbers and calculations are used to make sure that when your scanner says this is "light blue" (the number that represents light blue) your monitor understands exactly what "light blue" means after a calculation occurred that told it how the scanner renders the color and how that monitor should render the color. Same for your printer.

You do not have a positive reference in a color negative. How would you make a color profile for something that does not have a positive reference? A profile for something that's not there?

Bruce Watson
12-Dec-2006, 17:44
I am sure that Bruce & Kirk both know what they're about. But maybe it's easier stated in this most basic way: "you must get the scene information onto your film in a way that your scanner can see it".

That's the bottom line, yes.

Henry, if I were any good at writing/speaking I wouldn't need photography so much ;-)

Kirk Gittings
12-Dec-2006, 18:16
There is a LF photographer in the Midwest(I think), who does commercial drum scanning, who is known for his advocasy of scanning color negs. I can't for the life of me remember what his name is. I think his last name started with a B. Anyway I seem to remember some interesting info about scanning color negs on his website.

Kirk Gittings
12-Dec-2006, 18:24
That was Dan Burke, but my memory was wrong. Though he is a strong advocate of color negs, there is not in fact much info on his website.

Chris_Brown
13-Dec-2006, 11:41
OK, go do this and come back and explain how it worked for you.

Actually, there is a reason I mentioned scanning B&W negs in RGB mode, and it pertains to Imacon and PMT drum scanners. This may be moot to you, but others may find use for this info.

Both these scanners have independent capture elements for each color. The Imacon has a tri-linear CCD array, and PMT drum scanners have three photo multiplier tubes.

Since 2000, I've been using a Howtek drum scanner (http://www.aztek.com/) with Trident scanning software (http://www.colorbytesoftware.com/trident.htm) and it gives the option to scan in grayscale mode using one of the PMT channels. I tried it with each channel, but felt I was ignoring usuable data.

I don't print in grayscale mode on my printers (Epson, Canon) and usually apply a tint for content enhancement. To me, it makes sense to scan B&W negs in RGB mode.

As Caponigro says, "If you can see it, it's color." (http://www.johnpaulcaponigro.com/lib/downloads/technique/documents/tech_its_color.pdf)

Henry Ambrose
13-Dec-2006, 12:33
I didn't say you couldn't scan a B&W in RGB. For some older scanners this is a good way to get cleaner scans - its an old trick. But this has nothing to do with color profiles or the original question in this thread.

And I'm not putting down you or how you work - I was just trying to clear up the conversation and make the point that color profiles do not apply to B&W scanning.

I envy your scanner. That sounds like a great outfit. I bet you can make some great scans with it.

Kirk Gittings
13-Dec-2006, 15:38
It is even a good practice to scan in RGB with modern consumer scanners as one channel tends to be sharper and less noisy than the other two or a greyscale scan.

Stephen Sample
14-Dec-2006, 20:00
I started this question thread. There is a lot here to study and digest and it has been very helpful. I am getting pretty good 4 x 5 B&W negatives using the Steve Simmons Film Speed and Developing Time procedure. My contact print proofs are verifying this. I just received a refurbished Epson 4990 scanner and will be starting to scan those 4 x 5 B&W negatives using all of this advice. I also will make and use the input parameter controlled scene. Thank you for the seminar in scanning!

One more question about color positives. The scanner deal came with Silverfast SE, which Silverfast will update to the new version free of charge. This seems to be a quality outfit! I plan to learn the SE first. They will also give me an upgrade price on Ai and additional features. I do want to profile my scanner when I start scanning my 4 x 5 color Velvia 100F transparencies. How good and how good of a value is the Silverfast IT8 optional package? What are some other profiling packages that might do a good job and be cheaper?

Thanks again to everyone of you!

Kirk Gittings
14-Dec-2006, 21:16
My only experience with the Epson scanners was with the Monaco profiler and the Silverfast option. The SF was more accurate than the Monaco.