Color Matching Challenge: correct out of gamut color

[fike]

I have an unusual color matching problem.  I am trying to recreate the color in endsheets from a printed about a hundred years ago. We are trying to restore the book, but the endsheets are damaged. I can easily reproduce the graphical elements in PS, but getting a color match that is remotely close has been very problematic.

First I tried scanning and hand tweaking the color in ps.  That didn't work well.  I tried adding a gray card to the scan as a reference point. Still no dice....not even close.  Finally, I added my color checker to the scanned image and that got me closer, but I couldn't actually make a profile because the scanner only outputs TIFF and the color checker only works on real RAW files converted to DNG. 

So, I fell back to the last option which was to photograph the endsheet with the color checker sitting next to it and make a color profile.  Aside from problems preventing glare on the slightly shiny paper (I think I can deal with that by diffusing the light-sources better) I find that while the on-screen (calibrated device) display of the color to be as close as I have gotten, when I soft proof and check for out of gamut color, almost the entire image is out of gamut. When I print, it looks very dark and very brownish purple.  The sheet is closer to a rusty brown with a tint of purple in it.  I know, all those colors are challenging to work with.   To deal with the out of gamut color, I have been desaturating the offending channel (I think it is red).  This approach is obviously a problem, but I don't know how else to bring it in line so that I have a chance of printing a decent facsimile to a reliable paper profile.

This doesn't need to be exact. What I am trying to figure out is how to proceed to get a bit closer. 

Any suggestions...

[Mark D Segal]

I have no direct experience dealing with such a "challenge", but just thinking of options a bit: suppose you dropped the scanning route altogether, since it seems to be getting you nowhere. If you have a spectrophotometer and appropriate software, or a Pantone Color Cue, you should be able to read the colour value of the end-paper you are trying to emulate. Then in Photoshop open a new document and select a color fill for the background which has exactly the colour value you read with the spectro. Then texturize it, layer-in whatever other graphic elements you need (scanned and selected if need be), and perhaps you'll end-up with the effect you are looking for.

On the scanning options, I'd also be curious to know what scanning software you are using. Depending on that, it may be possible to bring the scan closer to your intended hue right at the scan stage, which would lessen the adjustment challenge in Photoshop.

[Victor Engel]

I've seen glare such as what you described eliminated in a printing press by using polarizers. The light illuminating the paper is polarized. The camera has opposite polarization.

[Ethan_Hansen]

If you have a Q-60 or other color reference, you could build a scanner profile. That might help.

If the problem is that the press or printer you are working with lacks the color range to reproduce the book colors, then you have a different problem. A good printer profile can help here - gamut mapping for out-of-gamut colors is handled better by some profiles than others. As a last resort, experiment with an Hue/Sat adjustment layer in Photoshop. It is the only PS tool that decouples saturation from hue and lightness. Carefully define the color range(s) to only include the problematic tints and first alter the lightness then the hue. You may well arrive at a color that is visually a better match to the original even if mathematically it is not. As a last resort you can lower saturation, but this is the cause of your muddy colors.

[fike]

Hmm...some good suggestions.

I don't have a spectrophotometer. I only have an i1 Display 2 for calibrating my monitor. 

Using a polarizer to tame the paper reflections...that is a good idea.  I will need to mess with that.

I am working with an Epson 2400 printer. I also have a 7880 that I could use too, but the gamut of these two devices is pretty similar, so I doubt switching over to the 7880 would be a game changer here.

I like the idea of somehow identifying the color with a spectrophotometer and then recreating the endsheet with a solid equivalent color in the RGB range.  It seems that the spec is the device I need to do this right, but I have always been under the impression that you don't get in to that type of hardware for less than a $1K.  I am not ready to spend that sort of money on this project yet.

I don't have a Q-60 and I don't have any profiling software.  I have not opened that end of the calibration can of worms yet.  Are there free or inexpensive profiling software packages available. I would love to profile my scanner because I have been having a heck of a time with my film scans.  the scanner really is the right tool for this job.

As for using hue/saturation adjustment layers...that is what I did to bring the out of gamut color channel (mostly red) under control.  It seems that in the process, I took too much color out and I am left with a very dark brown image. 

[na goodman]

If you can borrow a spectro to measure the color as Mark suggested I would do that. I had some restoration work to do and that was the only way I could get the color spot on.

[Tim Lookingbill]

With the DSLR route shooting Raw approach, have you tried changing your lights? Maybe different flash brands? direct sunlight? fluorescent lights made specific for photography?

An engineer told me a while back flatbed scanners in general have a narrower spectral bandwidth partially influenced by light source (usually fluorescent) compared to a DSLR which has a much wider gamut capturing capability and more options for better light sources.

However, from photographing a wide range of materials under various lights, even though I've found a DSLR will give more overall visually accurate results with some editing in post, the light source quality/characteristic and quantity combined with the camera's UV/IR filtering will skew the color on some objects.

I've gotten improved color accuracy from problematic objects that add their own spectral spike to the recipe by using different lights and/or adding more light by getting the object closer to the light or buying more light.

For instance I've gotten two different color renderings of the same object using the 18in. 5000K GE Sunshine and 18in. Philips 5000K Natural Sunshine tubes. Each will give their own slightly off WB cast to get the same accurate looking results meaning a WhiBal card may have to look subtly but noticeably red, blue or green tinted. The red spectrum from a flash puts so much red into this one particular object even though it's dark gray khaki yellow it doesn't even register in Photoshop as this color, but the surrounding scene looks normal and the object is hopelessly too red brown.

You may have to skew the WB just to get this color and use Selective Color editing to neutralize the WB surround that's not important or prominent in the image.

I'm grappling with this issue trying to make my Pentax K100D DSLR capture intense orange pomegranate flower lit by late afternoon sun shooting Raw keep from going over saturated. Exposure and WB doesn't fix it, but I photographed the flower indoors under the Philips fluorescent and it came out perfect.

So it's not just the device but the combination of the subject, filtering and light quality and quantity that have to be considered and changed. Light source is the easiest and cheapest and you can't do that on a scanner.

Here's what I'm talking about. Both are unedited default ACR previews.

[Mark D Segal]

I would love to profile my scanner because I have been having a heck of a time with my film scans.  the scanner really is the right tool for this job.

The scanner may not be the right tool for the job. It depends on whether you can get a nearly correct scan. (The other route I was suggesting above wouldn't need a DSLR either. It would be largely constructed in Photoshop.) Anyhow, turning to your film scanning issue, could you tell us what scanner and scanning software you are now using. That would be important to know in terms of recommending next steps.

[ErikKaffehr]

Hi,

If you are on Windows you could try Picture Windows Pro. It has a feature to match colors to color checker card.

http://dl-c.com/Temp/downloads/PW%20Doc/PW50.pdf (page 128 onward)

Best regards
Erik

I have an unusual color matching problem.  I am trying to recreate the color in endsheets from a printed about a hundred years ago. We are trying to restore the book, but the endsheets are damaged. I can easily reproduce the graphical elements in PS, but getting a color match that is remotely close has been very problematic.

First I tried scanning and hand tweaking the color in ps.  That didn't work well.  I tried adding a gray card to the scan as a reference point. Still no dice....not even close.  Finally, I added my color checker to the scanned image and that got me closer, but I couldn't actually make a profile because the scanner only outputs TIFF and the color checker only works on real RAW files converted to DNG. 

So, I fell back to the last option which was to photograph the endsheet with the color checker sitting next to it and make a color profile.  Aside from problems preventing glare on the slightly shiny paper (I think I can deal with that by diffusing the light-sources better) I find that while the on-screen (calibrated device) display of the color to be as close as I have gotten, when I soft proof and check for out of gamut color, almost the entire image is out of gamut. When I print, it looks very dark and very brownish purple.  The sheet is closer to a rusty brown with a tint of purple in it.  I know, all those colors are challenging to work with.   To deal with the out of gamut color, I have been desaturating the offending channel (I think it is red).  This approach is obviously a problem, but I don't know how else to bring it in line so that I have a chance of printing a decent facsimile to a reliable paper profile.

This doesn't need to be exact. What I am trying to figure out is how to proceed to get a bit closer. 

Any suggestions...

[Victor Engel]

Your flower is probably out of sRGB gamut. See http://forums.dpreview.com/forums/read.asp?forum=1003&message=38168773 for a discussion I started about a saturated rose (may be the most saturated flower I've seen).

[stefano]

For a relatively inexpensive spectro you can look at the ColorMunki - it has a specific mode to measure color. I have used it successfully many times, and it is as easy as the click of a button. That would give you a great start in Photoshop. For output gamut issues, I would start with perceptual intent and definitely a good profile for your printer/paper. Again, the colormunki can help you there - not as sophisticated as high end spectros, but it does produce some pretty decent profiles.

[Tim Lookingbill]

Your flower is probably out of sRGB gamut. See http://forums.dpreview.com/forums/read.asp?forum=1003&message=38168773 for a discussion I started about a saturated rose (may be the most saturated flower I've seen).

I'm going to have to disagree with you on that one.

It's not an out of gamut of sRGB issue. My camera doesn't capture Raw data in any color space. The rendering of such data may be mapped incorrectly to human perception definitions within a given set of default algorithms within a Raw converter. The scene may present a nonuniform spectral response not factored into these algorithms or else that orange flower doesn't even fit into ProPhotoRGB which I find that hard to believe.

This is why we "attempt" to profile our digital cameras as a way to more closely define the camera's color capturing characteristics which is influenced by the lens coating and UV/IR filter quality in front of the sensor as well as the A/D convertor behind the sensor. Sensors can be nonlinear and it's this nonlinearity that's not corrected for in the Raw converter algorithm that's causing that orange flower to blow out.

In other photography forums I'm picking up comments from photographers who say the X-rite Color Chart doesn't provide enough color samples to more closely define their camera's gamut capturing capabilities. This chart is primarily meant to get you in the ballpark based on established Raw converter default settings combined with what it knows about the camera sensor's nonlinearity which is pretty much zilch.

I can actually make the orange flower look normal in ACR which supports what I've mentioned above which I need to clarify is based on observing the behavior of several different digital capturing devices and what I've learned about color science and digital sensors.

I'm not a degreed professional.

[yannb]

Hello,

In the past I have created excellent scanner profiles using the Colorchecker SG chart, if you have access to it. You can do that with Profilemaker. Even the regular Colorchecker can be used, but it has far less patches of course.

Regards,
Yann

[Mark D Segal]

I can actually make the orange flower look normal in ACR

That means the raw data for doing so is there and it's a matter of settings.

[Tim Lookingbill]

That means the raw data for doing so is there and it's a matter of settings.

Right, which is what I indicated.

The settings which included Adobe Standard Beta 2 profile just happen to fix THAT particular exposed flower. Other exposures of other pomegranate flowers on the same tree lit by the same sun taken at different angles and exposures required completely different settings some of which included either severe WB adjust, HSL and/or severely contorted Point and Parametric curves combined with a custom SINGLE 6500K illuminant DNG profile based on the X-rite CC chart lit by mid day direct sunlight. The profile by itself couldn't fix the problem on its own.

I just read that dpreview thread and I don't even see a problem with their flower or any kind gamut issue.

[Tim Lookingbill]

Here's the results on another flower off the same tree and how different profiles will render each differently. Compare the top to the corrected one I posted previously and note that changing the WB to As Shot only slightly changed the color of the bark and the green leaves, but using a custom profile did the most damage while it worked just fine on the second flower but using curves instead of HSL panel.

[digitaldog]

In other photography forums I'm picking up comments from photographers who say the X-rite Color Chart doesn't provide enough color samples to more closely define their camera's gamut capturing capabilities. This chart is primarily meant to get you in the ballpark based on established Raw converter default settings combined with what it knows about the camera sensor's nonlinearity which is pretty much zilch.

I believe that is correct (the last part). The target here isn’t like the targets we used to build ICC camera profiles where its gamut affected the gamut of the resulting profiles. DNG profiles don’t operate this way in the processing chain is my guess (but we’d need Eric from Adobe to tell us for sure). Its like the old scripts we used to “calibrate” the ACR calibration pane using a Macbeth. The idea was, get a close match of those values as defined in this target.

[Tim Lookingbill]

The target here isn’t like the targets we used to build ICC camera profiles where its gamut affected the gamut of the resulting profiles.

That's interesting.

What targets are you talking about? What do they look like?

Talking with Brian Griffith, creator of Raw Developer for the Mac, says he has more elaborate processes in coming up with his default ICC based camera profiles but he never divulged what they were. His default previews were sometimes a bit more accurate on memory colors over ACR's on some images.

[digitaldog]

What targets are you talking about? What do they look like?

ColorChecker SG and DC (the later having a larger gamut due to the gloss patches that made shooting it awfully difficult).

[fike]

This is a very informative discussion. It seems like I could make a decent device profile with a known set of color patches, Q-60 for example, as long as I can include that color patch in the photo or scan of the subject. 

A couple fundamental questions:

• What software do I need to use to generate the actual profile? Is there a cheap alternative (cheap to me is less than $100, or free/open source)?
• Would this be an ICC profile or a camera raw device profile? Put another way, in what phase of processing (with what software) do I apply the profile?