How do scanners color correct C-41 negatives?

[Derek L]

Ok, lets start. Using the film Red Dmin as zero, then the Green is about 0.8 and the Blue is about 1.2. Taking this to the paper, the Red speed is zero, the Green is +0.8 log E faster, and the Blue is +1.2 log E faster in this example. Add a 50R to this and the G is 1.3 faster and the B is 1.7 faster. Thus, when you print it with a tungsten bulb with a 50R, the image is "nominally" neutral and correctly balanced.
PE

Hi PE, is it the case that if I shoot a grayscale with film and properly develop and RA-4 print it, the entire range of the grayscale will show as perfect neutrals (up to small errors inherent in the process)? I asked a friend of a friend at RIT and he indicated that this may not be the case (for instance the highlights may be warmer than neutral). But of course you have considerably more expertise, so I wonder if you could give a definitive answer.

If the entire range of neutrals does not print neutral, will a significant number of the midtones at least? Thank you.

 

[PhilBurton]

Guys,

As the "naive" person who posted message #51, I didn't grasp the complexity of all the issues involved. I appreciate all the responses. I guess I'll just start with scans of one particular film,and get that right. Then I'll go on to other films. Actually most of what I need to scan is not color negative. It's B&W negative, and various slide films.

 

[Alan Edward Klein]

Yes, I do!

What procedures do you follow, in post after the flat scans to get proper colors and exposure?

 

[Adrian Bacon]

One thing I don't understand is, if you've developed profiles for films to apply to scene-referred data, why not just skip the analog capture and use a profiled DSLR to get a scene-referred capture, then convert that? It's faster and cheaper.

I ask this half in jest—I understand there are properties other than color about film you might want to preserve. But you could, for instance, use any stock and make it look like any other. At this point you've erased all the color science that was engineered into the film and made it just another capture device. Don't get me wrong, that's technically impressive and I can't argue with the results, which I really like. It just strikes me as an odd way to go about scanning negatives.

My goal isn't to make all the emulsions look like each other. I can do that, and have, but I generally use the the generic c-41 profile, which is an average of all the totally zeroed out c-41 profiles that I've generated thus far. In effect, that generic c-41 profile is my digital "RA-4" paper. The result is a relatively neutral image that resembles what you'd get if you shot a DSLR and ran it through Adobe Camera Raw, however, as seen by the color patch charts I posted on the other thread, there is some variation with each emulsion. They all render similarly, but they are not the same. The only way I could have gotten there is by doing the work to profile and zero out all the emulsions so that I could get to an average. Now that I have those numbers for my particular setup, it's simple iterative refinement over time, which I've been at that stage for quite a while now, where I iteratively look for places where I could simplify or unify how I'm doing something and come up with a faster, or simpler, or easier to maintain way to do something. Some of that work has been get my hands on even more emulsions to zero out and add to the generic profile, some of it has been to support more/different hardware (I do upgrade the hardware I use over time), some of it has been to add handling different types of input, and some of it has been done because I'm just anal retentive that way. Everybody I process and scan film for generally comes away at least pretty happy with the results, because I'm fairly cost effectively giving them a pretty high fidelity color positive image that behaves just like the raw files from their digital cameras and pretty seamlessly fits right into their digital workflows that they already know and love, so for them, it's a chance to shoot film and get a little film aesthetic without any of the pain.

why not just skip the analog capture and use a profiled DSLR to get a scene-referred capture

Incidentally, all of my own digitally shot work is also run through my software. Not to diverge from this thread too much, but I primarily shoot Canon when shooting digitally, and Canon adds in this bias into their raw samples that from what I've seen from almost all raw converters except Canon's own DPP software, they totally mishandle the bias, which results in less dynamic range and more noise. People keep bashing Canon over dynamic range and image noise, and wonder why Canon doesn't do something about it.... Well... The truth is, performance-wise, Canon currently is behind its competitors, however, not by anywhere near as much as everybody says, and Canon can't make all the raw processors of the world do the right thing with their raw files. That bias is there for very good reasons which, depending on the application, makes Canon a superior camera compared to other brands. If you *correctly* handle the bias that Canon puts in their raw files, there's actually a whole *pile* of information that is effectively being discarded by most raw processors. This is actually the primary reason why I started developing Simple Image Tools. Using it to scan film is a secondary use.

 

[Ted Baker]

I agree that if you don't know the profile being used, then you can't do anything. But presumably the Fuji engineers know how the values in their custom, sRGB-like space can be converted to linear light measurements. If that is the case, there should be no problems with color managing their workflow in principle. (Whether they implemented such a feature is of course another question.)

I think we are mixing concepts here.

Taking a step back, in digital everything is essentially defined using a simple linear scale. Almost all manipulations that have scientific basis for example the Bradford method for adjusting white point the whole CIE xyz color model etc, sRGB is defined on the basis of that same linear model, but it is encoded using sRGB tone curve which is much more efficient if we are constrained by bandwidth. When doing a change fro sRGB to adobe RGB you need to convert to linear measurements, apply a matrix then re encode the data. Even a simple pixel resampling or sharpening operation must be done in this way... (otherwise you get weird colour shifts)

Now adding a tone curve in digital for aesthetic reasons has nothing to do with colorspace... When you open any RAW file in lightroom a tone curve is automatically applied... So any picture you open in lightroom is not RAW any more There is more to raw conversion that just demosacing.

However in color negative we use the gamma encoding for two purposes at the same time. One is to make better use of the available density range. We get many more stops this way. and the second purpose is using it in combination with the paper gamma or characteristic curve to get a tone curve that gives the desired pictorial result. i.e. the end result is the shadow and highlights are compressed and the midtones slightly expanded...

Your modelling MUST consider this last point.

 

[Photo Engineer]

Hi PE, is it the case that if I shoot a grayscale with film and properly develop and RA-4 print it, the entire range of the grayscale will show as perfect neutrals (up to small errors inherent in the process)? I asked a friend of a friend at RIT and he indicated that this may not be the case (for instance the highlights may be warmer than neutral). But of course you have considerably more expertise, so I wonder if you could give a definitive answer.

If the entire range of neutrals does not print neutral, will a significant number of the midtones at least? Thank you.

It may happen, but only if you have crossover - that is, assuming the exposure and lighting were correct you should have uniform neutrals unless the film has crossover..

PE

 

[Photo Engineer]

What procedures do you follow, in post after the flat scans to get proper colors and exposure?

None. The picture, as scanned, comes up correctly in PS.

PE

 

[Derek L]

I think we are mixing concepts here.

[...]

Your modelling MUST consider this last point.

I agree with the points you made. However, I don't understand why you believe they render it impossible to color manage scans from a Fuji scanner to render them substantially the same when printed, regardless of the printer used. (Of course, there are gamut mapping issues, but I already acknowledged those.)

For instance, the scanners are capable of giving sRGB jpeg output. With proper printer and paper profiling, I should be able to get relatively identical results with any printer that supports ICC profiles, since the properties of the sRGB color space are well known.

Could you identify what you think I'm overlooking?

 

[Derek L]

It may happen, but only if you have crossover - that is, assuming the exposure and lighting were correct you should have uniform neutrals unless the film has crossover..

PE

Thanks, this is great to know.

By the way, I finally found a used copy of that book by Evans you recommended to me for sale. It should be here in a few days.

 

[Ted Baker]

Could you identify what you think I'm overlooking?

Ok lets perhaps start with what you said here:

For instance, the scanners are capable of giving sRGB jpeg output. With proper printer and paper profiling, I should be able to get relatively identical results with any printer that supports ICC profiles, since the properties of the sRGB color space are well known.

Lets just for a start deconstruct this, except lets further simplifying it by discussing pure digital photography straight to screen and to a printer.

When you capture a photo with your DSLR the voltage created at the sensor which then becomes a number results in numbers that have a linear relationship to each other. For example a value of 36% is very close to twice as intense as 18%. (18% being roughly midscale in perception). (If the camera was perfect then 36% would be exactly double of 18%) The spectral sensitivity is carefully designed by the manufacture so that a colour matching function can then be used map those values to CIE XYZ or any other known colorspace. After capture the photo can then be processed directly in the camera to JPG, or you can do it with something like lightroom. The order of the steps is very similar, and the order DOES matter.

For example after demoscacing your data, you must first move the camera data to the new target colorspace, lets say you have chosen sRGB, so you must apply at least one forward matrix to move from camera colorspace to sRGB. Then you can apply the sRGB gamma function and then a tone curve (chosen by the manufacturer) to give the desired result.
Now at output stage the display hardware either by using appropriate phosphors in your CRT which match the primaries of the sRGB standard and using properties of the electron gun in the tube as the standard was envisaged to remove the sRGB gamma. Or you can use modern LCD or printer which remove the sRGB gamma from the signal and remap as appropriate, and apply an appropriate matrix/LUT to match the spectral qualities of the display or inks. The result after all this will still leave the tone curve that was added for pictorial reasons and any additional colour tweaks that are added by the manufacturer. That will distinguish one camera from another etc etc. Unless of course you believe all cameras are the same...

There are two points to make in this example which is broadly correct for illustration purposes. The first point is you can reorder a few things, but hopefully you can see that if you reorder things in the wrong way either the maths becomes very complicated or does not make sense or you get a different result...
The second point which is probably the most important is that there are something things in this pipeline that are added and then effectively removed so the net result is it makes no difference, leaving the manufacturer colour and tone tweaks... For example if your colors fit in the sRGB colorspace then it is as if didn't exist, otherwise it either clips them and you loose a few or it change all the colour slightly to fit the colorspace. But it doesn't change the signature of image, i.e. the tones and the general appearance of the colours etc. The manufactures tone curve and colour tweaks do!

In the discussion of colour negative films previously, hopefully you can see that there are several things in the camera negative, that "magically" disappear in the print, and leaving both the colour and tone "signature" designed by manufacturer. Keeping in mind the choice of print stock will vary the final result (there are still a handfull to choose) and also the exposure of both the camera negative and print will also vary the result. Many rate 400iso films at 100 or 200 etc. But there will be something that our brains can certainly distinguish.

Now after this long explanation if you profile as you describe with your 288 patches, how will you distinguish the characteristics that disappear when you print the camera negative from those that should be there which are the "signature" of the film?

What I think your are overlooking is that your modelling must consider at least part of how film works so that it contain at least some of its inherent characteristics. Otherwise you will remove all distinguishing characteristics from the film.

Hopefully this makes some sense!

 

[PhilBurton]

Thanks, this is great to know.

By the way, I finally found a used copy of that book by Evans you recommended to me for sale. It should be here in a few days.

What is the name of that book? I can't seem to find it in this message thread.

 

[Derek L]

What is the name of that book? I can't seem to find it in this message thread.

Principles of Color Photography by Brewer, Evans, and Hansen.

You can find more book recommendations in this thread: https://www.photrio.com/forum/threa...t-the-colorimetric-properties-of-film.169259/.

 

[Derek L]

Forgive me, but I am still slightly confused, so let me go through this step by step.

Ok lets perhaps start with what you said here:
Lets just for a start deconstruct this, except lets further simplifying it by discussing pure digital photography straight to screen and to a printer.

When you capture a photo with your DSLR the voltage created at the sensor which then becomes a number results in numbers that have a linear relationship to each other. For example a value of 36% is very close to twice as intense as 18%. (18% being roughly midscale in perception). (If the camera was perfect then 36% would be exactly double of 18%) The spectral sensitivity is carefully designed by the manufacture so that a colour matching function can then be used map those values to CIE XYZ or any other known colorspace. After capture the photo can then be processed directly in the camera to JPG, or you can do it with something like lightroom. The order of the steps is very similar, and the order DOES matter.

For example after demoscacing your data, you must first move the camera data to the new target colorspace, lets say you have chosen sRGB, so you must apply at least one forward matrix to move from camera colorspace to sRGB. Then you can apply the sRGB gamma function and then a tone curve (chosen by the manufacturer) to give the desired result.
Now at output stage the display hardware either by using appropriate phosphors in your CRT which match the primaries of the sRGB standard and using properties of the electron gun in the tube as the standard was envisaged to remove the sRGB gamma. Or you can use modern LCD or printer which remove the sRGB gamma from the signal and remap as appropriate, and apply an appropriate matrix/LUT to match the spectral qualities of the display or inks. The result after all this will still leave the tone curve that was added for pictorial reasons and any additional colour tweaks that are added by the manufacturer. That will distinguish one camera from another etc etc. Unless of course you believe all cameras are the same...

I agree with all this 100%.

There are two points to make in this example which is broadly correct for illustration purposes. The first point is you can reorder a few things, but hopefully you can see that if you reorder things in the wrong way either the maths becomes very complicated or does not make sense or you get a different result...
The second point which is probably the most important is that there are something things in this pipeline that are added and then effectively removed so the net result is it makes no difference, leaving the manufacturer colour and tone tweaks... For example if your colors fit in the sRGB colorspace then it is as if didn't exist, otherwise it either clips them and you loose a few or it change all the colour slightly to fit the colorspace. But it doesn't change the signature of image, i.e. the tones and the general appearance of the colours etc. The manufactures tone curve and colour tweaks do!

Agreed.

In the discussion of colour negative films previously, hopefully you can see that there are several things in the camera negative, that "magically" disappear in the print, and leaving both the colour and tone "signature" designed by manufacturer. Keeping in mind the choice of print stock will vary the final result (there are still a handfull to choose) and also the exposure of both the camera negative and print will also vary the result. Many rate 400iso films at 100 or 200 etc. But there will be something that our brains can certainly distinguish.

Here's where you lose me. If by things that "magically" disappear in the print you mean the orange mask, for example, I agree.

Now after this long explanation if you profile as you describe with your 288 patches, how will you distinguish the characteristics that disappear when you print the camera negative from those that should be there which are the "signature" of the film?

What I think your are overlooking is that your modelling must consider at least part of how film works so that it contain at least some of its inherent characteristics. Otherwise you will remove all distinguishing characteristics from the film.

Hopefully this makes some sense!

So, my proposal is to profile based on the scans returned to me by the lab and the original negatives. I'd turn my DSLR into an ersatz densitometer and profile the entire negative -> scan process as a black box. I agree this includes both what's baked into the film by Kodak *and* whatever transforms are made by the Fuji scanner. But, if I want Fuji scanner output, it'll get me pretty close to that.

By the way, after some more thinking based on PE's description, I think simpler approaches would work just as well. But in principle, if I all I care about is profiling the entire film+scanner imaging chain, I think the 288 target would get very good results. It's just overkill.

 

[MattKing]

f by things that "magically" disappear in the print you mean the orange mask, for example, I agree.

And more importantly, the inversions of the colours themselves!

 

[Adrian Bacon]

By the way, after some more thinking based on PE's description, I think simpler approaches would work just as well. But in principle, if I all I care about is profiling the entire film+scanner imaging chain, I think the 288 target would get very good results. It's just overkill.

My original implementation (the one with all the Macbeth charts for each emulsion in the other thread) is based on 6 patches. I generate 360 adjustments (one for each whole hue degree), but the waypoints where the measurements are taken are the red, green, blue, yellow, magenta, and cyan patches. My newer method that I'm working out now, uses 16 patches for the measurement waypoints, and it won't render significantly different from the old way, just some of the hues and saturations will be a bit more accurate and less prone to pushing one way or the other. Charts with a large number of patches are good if you want to map out and characterize a given sensor/emulsion, but, to get reasonably acceptable pictorial results, you actually really only need a handful of patches that are placed in the right spots.

 

[Ted Baker]

So, my proposal is to profile based on the scans returned to me by the lab and the original negatives. I'd turn my DSLR into an ersatz densitometer and profile the entire negative -> scan process as a black box. I agree this includes both what's baked into the film by Kodak *and* whatever transforms are made by the Fuji scanner. But, if I want Fuji scanner output, it'll get me pretty close to that.

OK I see your proposal more clearly. That might work if you want to scan that same piece film again again and again or you the same type of film and expose it the same EI and the contents of the scene is essentially the same and you are happy with the "exposure" used for the print. You can use a HaldCLUT for this, or there is something similar you can load into photoshop.

Here's where you lose me. If by things that "magically" disappear in the print you mean the orange mask, for example, I agree.

Well there is the mask, but there are things that the DSLR sensor will pick up that the paper won't. The paper has a blind spot around 600um, there is the gamma in the negative that as discussed previously both allows the density range to be better used but it also contains "pictorial" information. The "pictorial" aspect need to remain. etc etc. You need to know which is which.

Various fudges are acceptable it depends on how faithfully you wish to model the process. For example many are very happy with the results from an epson scanner using epson scan, output into sRGB colorspace etc etc. Some are very skilled in photoshop and get excellent results that way.

OR as an idea you could...

create a negative with enough colour samples across the entire useable density range along with a white, perhaps with a film recorder or optically with a camera. You could measure each of these with status-m (just for fun) and you could then scan them with your camera, with a setup that you can duplicate exactly. i.e. lightsource, exposure etc.

You could then print that negative on your chosen print stock, balanced with filtration or color timing to the give correct neutral tone on your white patches through the essential midtones. You could then measure all your patches with spectrometer on your print and (status-a just for fun again). You also need to repeat this a few times for different print exposures.

Now you have some excellent data to work with...

 

[Derek L]

I think we agree on everything, then.

I would love to use a spectrometer to profile a RA-4 print of a large color reference chart, but unfortunately I can't find anyone online who will optically print RA-4 for me (and I don't have the materials or expertise to do it myself). I also don't own a spectrometer, though I suppose I could rent one and the cost wouldn't be so bad.

 

[Alan Edward Klein]

None. The picture, as scanned, comes up correctly in PS.

PE

What do you mean that they come up correctly in PS?

 

[Photo Engineer]

The negatives I scan come up fully balanced and with the correct tone scale in Photo Shop! I can mix negatives and slides in one scan, sometimes the same picture and they are just fine.

Ok?

PE

 

[Alan Edward Klein]

The negatives I scan come up fully balanced and with the correct tone scale in Photo Shop! I can mix negatives and slides in one scan, sometimes the same picture and they are just fine.

Ok?

PE

How do you scan flat to get them already adjusted? You must have the scanner on auto something.

 

[Photo Engineer]

I do nothing to the Epson software. I merely point PS at the scanner and click scan! You can view the results in my gallery. All of those were scanned the same way. Including slides, internegatives and negatives. Some of the slides were interpositives.

PE

 

[Alan Edward Klein]

I do nothing to the Epson software. I merely point PS at the scanner and click scan! You can view the results in my gallery. All of those were scanned the same way. Including slides, internegatives and negatives. Some of the slides were interpositives.

PE

Your Epsonscan is apparently in an Auto mode that corrects the scan. It's not scanning flat. I believe Auto is the default mode. You have to select No Color Corrections in the Configurations Tab to scan flat. Then you can make all adjustment in Post Processing. I use to scan in Auto. But I found that it was blowing out the highlights. To prevent that, I had to adjust levels manually. You can see your settings if you open Epson scan. There are also four standard modes, Pro, Auto, Office and Home. I use Pro mode, but you still have to select Auto, or No Color Corrections or another setting in this mode as well.

 

[Ted Baker]

I think we agree on everything, then.

A fudge is not necessarily wrong, but if your at the level of writing software to do this task, hopefully this thread is illuminating. I am just an enthusiast like the half dozen or so people on this forum that are attempting to reinvent the wheel out of necessity. I spent a few years printing RA-4 (some professionally), so doing some of the above profiling as described was not too difficult.

This technical document on the ARRI version of the Kodak Cineon system is well worth a read and should help your further your knowledge. Interesting in 2005 they profiled using 1728 patches in a simlar method as previously described i.e. 123 LUT but this is after all the other modelling.

http://dicomp.arri.de/digital/digital_systems/DIcompanion/index.html

and this diagram which describes one way of doing the signal processing:

I have not found anything from Noritsu or Fuji with any of this level of detail revealed. It is a pity that Kodak did not release more of there own stuff. I think the Pakon source code is lost for example, and PhotoCD is totally defunct... PhotoCD had something called film terms which is a proprietary way of storing LUTs which characterised all Kodak negative films of the day and many others as well... All of that is probably lost. I believe the aforementioned Giorianni invented PhotoCD.

 

[Photo Engineer]

Your Epsonscan is apparently in an Auto mode that corrects the scan. It's not scanning flat. I believe Auto is the default mode. You have to select No Color Corrections in the Configurations Tab to scan flat. Then you can make all adjustment in Post Processing. I use to scan in Auto. But I found that it was blowing out the highlights. To prevent that, I had to adjust levels manually. You can see your settings if you open Epson scan. There are also four standard modes, Pro, Auto, Office and Home. I use Pro mode, but you still have to select Auto, or No Color Corrections or another setting in this mode as well.

Do you think I am unaware of these things?

The entire pop up window for the scanner is there when I use it, but I don't have to change the options. It works just fine as is.

PE

 

[Photo Engineer]

Ted, the curve of any internegative or reproduction of a negative cannot be linear, if I understand what you have said.

PE