Problem: Copying Artwork & Matching Colours

[William Walker]

Hi

I have just had a frustrating day trying to make a print from a painting.

1. I lit the painting with studio lights, using a Color-Checker Passport generated profile and a graycard for white balance.
2. I then made a few small adjustments, with the painting still illuminated, on my calibrated NEC PA272W. I made an accurate on-screen match with the painting.
3. I then soft-proofed the picture and made a print using a custom-made profile for Hahnemuhle Museum Etching on a Canon iPF8400.

There is a Bluish-green colour (jade, according to the artist) that I cannot even get close to - it is miles out! It prints to a dark blue no matter how I try to tweak it...

At some point though, it becomes ridiculous in terms of what a proper colour-managed workflow should be!

The two pictures give you an idea of the problem.

My question is: Can someone pick up something blatant that I have missed, or, are some colours just impossible to reproduce?

PS. I have no similar issues when printing "normal" photographs - the colours I see on my display are pretty-much the colours I get on the print.

Any ideas will be welcomed!
Thanks
William

[Czornyj]

This color should be a pice of cake for iPF8400, so apparently something is very wrong here. How did you create custom profile? What rendering intent are you using? Are you using Photoshop Print-Plug in, or printing from driver? PC or Mac?

[Bart_van_der_Wolf]

Hi

I have just had a frustrating day trying to make a print from a painting.

1. I lit the painting with studio lights, using a Color-Checker Passport generated profile and a graycard for white balance.

Hi William,

Maybe it has something to do with the profile you used as a base profile, e.g. Lightroom and ACR can make use of the Adobe DNG Profile Editor, but it is just a modification of a base profile, and we do not know how good that base profile was for the specific color you are having trouble with. Capture One can also create 'modified' (ICC) profiles with its built-in Color Editor tool.

Building a profile from scratch should give a better chance of success, especially if patches of the difficult color can be added to the profile creation/optimization.

It may also be a result of the camera's native color sensitivity as seen through the CFA. A bluish/green jade color, may well fall exactly between the Green and Blue transmission curves, and it depends on the amount of overlap of both color ranges if it stands a chance of being picked up with good reconstruction quality. So a different camera might in that case solve part of the issue.
 

2. I then made a few small adjustments, with the painting still illuminated, on my calibrated NEC PA272W. I made an accurate on-screen match with the painting.
3. I then soft-proofed the picture and made a print using a custom-made profile for Hahnemuhle Museum Etching on a Canon iPF8400.

We'll have to assume those profiles were good. You may want to check that by printing some critical test images, or a Granger rainbow.

There is a Bluish-green colour (jade, according to the artist) that I cannot even get close to - it is miles out! It prints to a dark blue no matter how I try to tweak it...

At some point though, it becomes ridiculous in terms of what a proper colour-managed workflow should be!

Well, most camera profiles are created for pleasing colors, not accurate ones. You need something more accurate than pleasing.

Cheers,
Bart

[Mark D Segal]

I agree with Marcin that the extent of disconnect shown between the print and the original is egregious. Whether it should be a "piece of cake" or not - in any inkjet printer - is another question. The spectral properties of paints - how they are lit and how they reflect light are known for their ability to bedevil our standard profiling approaches, even with very good profile making equipment.

I think it's necessary to begin the remedial work from the lighting of the painting onward. The colour temperature of the lighting could be influencing the problem, notwithstanding that you use a gray patch for neutralizing the lighting. When you use a gray patch to neutralize a colour cast caused by the lighting, it may not be readjusting all the other colours accurately, perhaps partly on account of the properties of the paint colours - and this would happen well before the printing stage. This readjustment happens in your raw converter, so the choice of camera profile, or perhaps the creation of a custom camera profile may be a useful precaution.

For the print stage, there is considerable discussion on the Internet about creating profiles for dealing with the accurate reproduction of paintings and drawings. It may be useful to consult that literature. Recall that every museum and publisher dealing with the archiving or dissemination of photographs of paintings, whether oil or watercolours, needs to deal with this problem. They manage it more or less successfully - some very successfully with a lot of post-processing. But I agree with Peter that "compromise" is often the state of the art and science. It is very often necessary to intervene with targeted adjustments of specific colours in the photo editing stage to get everything to reproduce to an acceptable degree of perceptual similarity with the original - do not count on profiling alone as a complete solution. Soft-proofing is essential to minimize waste and frustration.

[Doug Gray]

Reproduction work is some of the most demanding technically, and least demanding artistically, processes in color capture/printing.

To emphasize what has been said already, the most critical aspect of repro work is that the captured image should be processed using a "scene referenced" profile. Either an ICC profile or an Adobe DCP. Also, it's very difficult to do repro work without a spectro to measure the accuracy of critical colors on the print v those on the original.

[William Walker]

This color should be a pice of cake for iPF8400, so apparently something is very wrong here. How did you create custom profile? What rendering intent are you using? Are you using Photoshop Print-Plug in, or printing from driver? PC or Mac?

Thank you all for your replies, let me try and answer a few questions:

1. I created the profile in Lightroom using the X-Rite Preset - Export to ColourChecker Passport. (See Pic).
2. Once the profile is loaded I reset/check white balance.
3. Rendering intent - Perceptual - it looked better than Relative.
4. All this is done through Lightroom on a MacPro with NEC display.

The section of "jade" actually shows "out-of-gamut" during soft-proofing, but disappears after a -26 reduction of Aqua Saturation - makes no difference.

Mark - I take your point about having to go in and do targeted adjustments, but like Marcin, I also cannot believe that the iPF8400 is not capable of producing that colour!

[Mark D Segal]

Mark - I take your point about having to go in and do targeted adjustments, but like Marcin, I also cannot believe that the iPF8400 is not capable of producing that colour!

Looks like a head-banger! Can you try this test: read several characteristic spots of the troublesome colour off the original with a spectro; read the LAB values of the colours you sampled, then go into Photoshop this time and create patches having those LAB colours, save them as TIFF files, softproof them with your printing profile and see what emerges - check with Relative, Perceptual and Absolute RI just in case they are OOG, to see which handles them best. Are you using ProPhoto colour space for all your work? if not, there is more risk of saturation clipping.

[Czornyj]

4. All this is done through Lightroom on a MacPro with NEC display

How did you print targets to create the profile? Are you aware of Canon iPF driver issue with ColorSync? Did you use Doyle Yoder's patch to fix it?
http://www.dypinc.com/Canon/AppColorMatchingInfo.xml.zip

[Mark D Segal]

Adobe Color Print Utility downloadable free from the Adobe website as well.

[Czornyj]

Adobe Color Print Utility downloadable free from the Adobe website as well.

Due to ColorSync + iPF driver issue LR doesn't turn off ColorSync conversion, so when you manually select ICC profile in LR Print module, double color management occurs. You need to apply Doyle's patch to fix this issue.

This could also affect targets printed to create custom HGE profile, that's why I'm asking about the way they were printed.

[Mark D Segal]

Hi Marcin, yes I know what the issue is. I was just suggesting that an alternative way of printing profiling targets without risk of interfering colour management is to use the Adobe Color Print Utility - it was designed for that and at least it works fine from Epson printers, as it did when I used it for Canon's new Pro-1000. Is there a reason why it wouldn't work as well for the previous generation of Canon professional printers?

[Czornyj]

The iPF PRO series drivers are fixed by Canon.

As for ACPU and iPF x400, I'm just not certain - I fixed the issue with Doyle's patch years ago, and always printed targets from PS Print Plug-in (which is more convenient than ACPU).

There's a chance that the target was printed the right way, so the OP should just apply the Doyle's patch and everything should get back to normal. Otherwise I'd use stock Hahnemühle profile, or Canon profile for textured heavyweight fine art paper.

[Mark D Segal]

OK thanks.

Actually ACPU is pretty easy to use - same basic driver settings as for regular printing. It eliminates colour management as a concern for printing profiling targets. I've found it really reliable. You may recall the history.

[Doug Gray]

I doubt the problem is on the printer side. It sounds more like the photographed image is being processed using the normal Output Referenced process. For artwork reproduction you need a Scene Referenced (except for white point) DCP or ICC and the image has to start off as a RAW image. Camera images that are generated as jpegs are virtually always Output Referenced and have S-curves applied for dynamic range reduction. This will never produce good images for reproduction.

[Mark D Segal]

I doubt the problem is on the printer side. It sounds more like the photographed image is being processed using the normal Output Referenced process. For artwork reproduction you need a Scene Referenced (except for white point) DCP or ICC and the image has to start off as a RAW image. Camera images that are generated as jpegs are virtually always Output Referenced and have S-curves applied for dynamic range reduction. This will never produce good images for reproduction.

OP doesn't say directly, but some information suggests he probably works with raw files. The intriguing thing that points to printing as the issue was his statement that he got a good match between the softproofed photo on his display and the painting itself. This would seem to suggest that the capture end of the process may be OK.

[Doug Gray]

OP doesn't say directly, but some information suggests he probably works with raw files. The intriguing thing that points to printing as the issue was his statement that he got a good match between the softproofed photo on his display and the painting itself. This would seem to suggest that the capture end of the process may be OK.

True. But he also says his normal printing works fine. Hard to know what's going on. We really need more specific information on his workflow and settings from the camera on. It sure looks like he isn't using scene referenced processes. And he may have printer workflow issues on top of that. With a reasonable setup you can get pretty good results doing repro without even looking at a monitor image. Even though cameras are not particularly good colorimetric devices they do have excellent linearity in RAW mode and reasonably good color. But I've never seen blue colors as far off as he's getting.

What color space are the linked images in? They are untagged so I assume sRGB. Cyan and teal in sRGB are well within the gamut of virtually any printer.

[Mark D Segal]

True. But he also says his normal printing works fine. Hard to know what's going on. We really need more specific information on his workflow and settings from the camera on. It sure looks like he isn't using scene referenced processes. And he may have printer workflow issues on top of that. With a reasonable setup you can get pretty good results doing repro without even looking at a monitor image. Even though cameras are not particularly good colorimetric devices they do have excellent linearity in RAW mode and reasonably good color. But I've never seen blue colors as far off as he's getting.

What color space are the linked images in? They are untagged so I assume sRGB. Cyan and teal in sRGB are well within the gamut of virtually any printer.

Yes, I also raised the colour working space issue in Reply #7. If indeed they are in sRGB, it's quite conceivable that the troublesome part of the colour spectrum is OOG for sRGB, and if that's the case, then depending on his RI, the OOG colours would be remapped to "something", and it is that "something" which bears such poor resemblance to the original artwork. But you are quite correct we need more information - this is only suggestive and hypothetical.

[Doug Gray]

Yes, I also raised the colour working space issue in Reply #7. If indeed they are in sRGB, it's quite conceivable that the troublesome part of the colour spectrum is OOG for sRGB, and if that's the case, then depending on his RI, the OOG colours would be remapped to "something", and it is that "something" which bears such poor resemblance to the original artwork. But you are quite correct we need more information - this is only suggestive and hypothetical.

Indeed. Those are all good points.

[Tim Lookingbill]

I just did a test conversion of his sRGB? top reference image (the one the OP says looks exactly like the original painting) converting to numerous coated/uncoated web press profiles, Frontier Dry Lab profiles using RI, AbsCoL and Perceptual intents and they all reproduced the jade color without a problem considering Soft Proof works as it should.

The jade in the reference sRGB image has a clipped red channel to zero, converting to AdobeRGB shows no clipped numbers. I bet you my $50 Epson All In One using Printer Manages Color on my Mac could reproduce that jade easily. Something is really screwed up with the profile or how it's communicating with the Canon printer driver.

Like Marcin suggested try using the downloadable paper profile or a canned profile to see if it can reproduce the jade color.

[Mark D Segal]

Something is really screwed up with the profile or how it's communicating with the Canon printer driver.

Like Marcin suggested try using the downloadable paper profile or a canned profile to see if it can reproduce the jade color.

Tim, if you go back to William's originating post, Step 3, the softproof generated from his custom profile works at least on his display. Not ruling out the possibility that his B>A and A>B tables in the profile are behaving inconsistently (i.e. as you suggest, screwed-up profile), but I wonder how likely. In any event the suggestion to try using a canned profile would indeed be a good check for that.