Difficulty Dupliating Flat Artwork

[Craig Stocks]

I'm sure with the experience here someone else has solved this problem, but so far, the solution is eluding me.

My challenge is photographing a watercolor painting with very subtle colors. The problem is getting the colors to match the original, they always photograph just a little light and de saturated.

I'm also working with a bold acrylic painting it matches perfectly, it's just the watercolor that is trying my patience.

My process:

Photographed with Phase One XF / IQ4150 and Phase One 120 macro. I've also tried my Sony a7r2. Light is from two Alien Bee strobes set at 45 degrees.

I photograph a year-old Xrite ColorChecker as a calibration and profile target, then photograph the artwork. I also photograph and apply at LCC frame. Both frames receive identical processing based on the ColorChecker.

Color process: I've tried both the Phase One Flash - Flat Artwork profile and I've created my own duplication profile with Lumariver Profile Designer. I select the Linear Response curve. Generally, if properly exposed the gray squares fall almost exactly at their published values.

I make very small tweaks using the Luma Curve and small color adjustments using the Color Editor seeking a visual match with a ColorChecker overlay with half of each square transparent. I compare to both the published standard values and measurements of my own ColorChecker.

I then export the image using Adobe RGB 1998 (I've also tried ProPhoto RGB)

I open both frames in Photoshop and again check the tones and colors. If needed, I tweak them into alignment based on both a visual overlay and the RGB numbers.

I've tried strip lights and simple reflectors, both in pairs and singly to see if shadows made the difference. I've also added UV filters to the lights, polarizing filters, both UV and polarizing filters on the lights and cross-polarizing. (I did find that adding the UV & polarizing filters kept the yellows from going too peach colored.)

Needless to say I've spend WAY too much time trying to find a combination that gave me a match on the ColorChecker and the artwork but finally gave up and tweaked the artwork photo to achieve a match on my calibrated Eizo as well as when printed on my Epson 9900.

So, what am I doing wrong? What do you do when the calibration target matches and the artwork doesn't?

I've attached a couple comparison frames showing the difference. The photo of the ColorChecker has a reference overlay where the bottom-left half of each square shows the target color. The smaller center square shows my measured value for my ColorChecker and the top-right portion is transparent and shows the ColorChecker in the photo. This example was photographed with polarized and UV filtered strobes (not cross-polarized) and processed with the Phase One Flash - Flat Art profile and Linear Resposne curve and an LCC was applied. White balance set from the second brightest square. No other tone or color adjustments adjustments.

 

[Boinger]

This may be a silly question but what is the artwork backed on?

How thick is the paper in question?

Something that comes to my mind is the light may be passing through the paper to small degree and thus you are losing saturation. Try putting the artwork on a solid pure white thick backing, and see if that makes a difference?

 

[dchew]

Craig,
I went through almost exactly the same thing when I was photographing 200+ samples of leather dyed 200+ different ways. I too tried both standard profiles and custom profiles. Nothing worked perfectly, and every single sample included one colorchecker image. I think Doug will chime in here and point out the detailed workflow and lighting that is required to get everything perfect when doing critical copying / archiving.

I did find the "LeafLF5 LF5 Product" profile in the Leaf Color IQ 100MP folder to be one of the more consistent, even though it is for the wrong back. You might try that since it is so quick and easy to test. Bright reds were off, but the orange/yellow and blue/greens were pretty good with that profile applied to the IQ4 150 images.

Dave

 

[Craig Stocks]

This may be a silly question but what is the artwork backed on?

How thick is the paper in question?

Something that comes to my mind is the light may be passing through the paper to small degree and thus you are losing saturation. Try putting the artwork on a solid pure white thick backing, and see if that makes a difference?

The artwork is pretty opaque and was sitting on a white matboard. Since it is coming out too light maybe I should try putting it directly on a dark background.

 

[Boinger]

The artwork is pretty opaque and was sitting on a white matboard. Since it is coming out too light maybe I should try putting it directly on a dark background.

That might actually work better. See the colorchecker palette has a dark backing doing the same might have the same effect on your artwork.

What I am guessing is happening is that the light is passing through and losing the saturation. Let me know how the dark backing works out?

 

[Craig Stocks]

One theory tested and discarded. I photographed on a split background (half white - half black) and no difference.

I'm beginning to believe that watercolors just photograph differently. That or pastel colors are very problematic overall.

 

[faberryman]

Sensors, including the IQ4150, are not perfect. Just imagine if you were using film.

 

[pinktank]

Technically speaking, if you were shooting for a museum or something, you would ideally have an object level target that would remain with the archival image and not do any adjustments. That being said, some points:

1) How are you setting your initial exposure?
2) When you photograph the colorchecker, do it on a black background and not on top of the watercolor to reduce glare/internal reflections. Also make sure that light is not bouncing your lens from somewhere else
3) What light are you 'proofing' under? Your monitor shouldn't really match a paper object, one is on paper lit with incident light, the other is backlit, they have different surrounds, and many other such reasons why our humans eye are not good at judging whether the capture is 'accurate' from a display
4) Similarly, how and with what settings is your printer profiled at? Again, are you printing on the same paper that the watercolor was printed on, or at least a visual match? Are you printing with the same margins and trimming the paper-white section?

 

[dougpeterson]

On a plane about to take off so excuse brevity...

https://dtculturalheritage.com/product-category/digitization-guides/

The reflective digitization guide covers flat art reproduction in Capture One. It focuses mostly on Capture One Cultural Heritage but most of the lessons and concepts can be applied in the standard version of C1.

That said, art repro and color are both science and art. There are numerous situations in which strict adherence to a numerically accurate workflow won’t address a specific need. Art repro shouldn’t be exclusively “paint by numbers”.

Notably, over the last couple years, many CH institutions have switched or are planning to switch, from strobes to the DT Photon as it allows real-time proofing, WYSIWYG, and the best spectral accuracy and smoothness you can get.

 

[Craig Stocks]

Technically speaking, if you were shooting for a museum or something, you would ideally have an object level target that would remain with the archival image and not do any adjustments. That being said, some points:

1) How are you setting your initial exposure?
2) When you photograph the colorchecker, do it on a black background and not on top of the watercolor to reduce glare/internal reflections. Also make sure that light is not bouncing your lens from somewhere else
3) What light are you 'proofing' under? Your monitor shouldn't really match a paper object, one is on paper lit with incident light, the other is backlit, they have different surrounds, and many other such reasons why our humans eye are not good at judging whether the capture is 'accurate' from a display
4) Similarly, how and with what settings is your printer profiled at? Again, are you printing on the same paper that the watercolor was printed on, or at least a visual match? Are you printing with the same margins and trimming the paper-white section?

I've been using a ColorChecker as a poor-man's object level target but it wasn't helping me so I quit using it as I explored options. When it was included it showed essentially the same values as the large ColorChecker.

1 - Exposure is set so that the gray squares match the publshed luminosity when processed with a flat art profile and linear curve.

2 - I haven't tried that. But, even if I'm getting glare it should affect both the ColorChecker frame and the art frame

3 - I've used a variety of LED, tungsten and daylight, always the same result.

4 - Printing was on a variety of papers including Epson Premium Luster, Brilliant Glossy and Canson Baryta, always the same result. I used factory profiles for the Epson and Canson papers and my own profile for the Brilliant paper.

My goal here isn't to produce prints, they were simply a convenient way to get another objective (?) view of the result and try to sort out the differences. Basically, the images of the ColorChecker match and photos of a colorful acrylic painting match all match the original. It's only the watercolor that doesn't want to match.

 

[Craig Stocks]

On a plane about to take off so excuse brevity...

https://dtculturalheritage.com/product-category/digitization-guides/

The reflective digitization guide covers flat art reproduction in Capture One. It focuses mostly on Capture One Cultural Heritage but most of the lessons and concepts can be applied in the standard version of C1.

That said, art repro and color are both science and art. There are numerous situations in which strict adherence to a numerically accurate workflow won’t address a specific need. Art repro shouldn’t be exclusively “paint by numbers”.

Notably, over the last couple years, many CH institutions have switched or are planning to switch, from strobes to the DT Photon as it allows real-time proofing, WYSIWYG, and the best spectral accuracy and smoothness you can get.

Thanks Doug. Actually I purchased and read the guide some time ago and found it to be an excellent resource. I didn't really find any steps in my process I needed to change but it did help me understand more of what's happening, such as the difference in the two linear curves.

I appreciate your statement that art repro shouldn't be exclusively "paint by numbers" but it bothers me when there's a break in the process. In this case adjusting for the artwork destroys the validity of the target.

I would suspect that my in situ Lumariver profile was to blame but I get the same result using the built in Phase One flat art profile.

 

[Jorgen Udvang]

It's been a long time since I did this kind of work, but what I learned back then was that symetric light sources made flat artworks look "flat" and desaturated. Even acrylic paint and most certainly art paper have some structure that need to be emphasised by asymetric lighting, creating a "micro shadow" effect that boosts the dark tones that lie hidden even in the lightest pastels. My gear at the time was rather basic compared to what you are using, and it took me a lot of experimenting to get it right, but the principle should still be the same.

Since the ColorChecker probably has a different surface structure than the art paper you are working with, my guess is that the colour response will also change differently when using angled light sources, and although the variations may be subtle, it might be noticeable with this kind of work.

 

[Craig Stocks]

It's been a long time since I did this kind of work, but what I learned back then was that symetric light sources made flat artworks look "flat" and desaturated. Even acrylic paint and most certainly art paper have some structure that need to be emphasised by asymetric lighting, creating a "micro shadow" effect that boosts the dark tones that lie hidden even in the lightest pastels. My gear at the time was rather basic compared to what you are using, and it took me a lot of experimenting to get it right, but the principle should still be the same.

Since the ColorChecker probably has a different surface structure than the art paper you are working with, my guess is that the colour response will also change differently when using angled light sources, and although the variations may be subtle, it might be noticeable with this kind of work.

I’ll have to try asymmetrical lighting again. I tried a couple different variations already and had pretty much the same result. I did not work through different lighting ratios so that may be what’s missing.

 

[RLB]

Accurately digitizing flat art is vastly more difficult than one might initially assume. There are a myriad of variables from DB, to lenses, to lighting issues to reference targets, monitor types, calibration, profiles, curves, processing workflow, did I mention the art itself? I've been digitizing art for professional for museums for more than 20 years, and assure you that to do this with great accuracy is huge challenge for anyone even with great equipment.

I'd like to be of help but that would be like teaching someone to be a great surgeon by email. Far to may variables to even start.

 

[pinktank]

It cannot be 'the same' between all the light sources and the papers you mentioned, they are by nature very different. If you are not lighting it the same way as the photograph (45 degrees crossed) in comparing it to the screen, the watercolor pigment/paper combination might be responding differently to that as well. By the 'same' I'm guessing you mean that the other painting looks similar and the watercolor dissimilar. That is also about your eyes having an easier time matching a fully colored-in/toned object.
When you have artwork that contains a lot of lighter tones revealing large areas of base, it becomes hard to reproduce it in a different medium. It is easier when the frame is filled with middle tones like that of the acrylic painting, partly because they obscure the media white/blank and our eyes/brains have an easier time with that.

 

[Craig Stocks]

Accurately digitizing flat art is vastly more difficult than one might initially assume. There are a myriad of variables from DB, to lenses, to lighting issues to reference targets, monitor types, calibration, profiles, curves, processing workflow, did I mention the art itself? I've been digitizing art for professional for museums for more than 20 years, and assure you that to do this with great accuracy is huge challenge for anyone even with great equipment.

I'd like to be of help but that would be like teaching someone to be a great surgeon by email. Far to may variables to even start.

I guess I'm not sure why you took the time to post a reply just to say that you can't help. Are you suggesting it's so difficult that there's no hope of ever learning and I should simply give up?

 

[kdphotography]

There are little things you can do (small investment) to improve your equipment used and offer greater consistency. For example, the jump from Alien Bees strobes to Einsteins would be a very minor investment, but the color consistency from the Einsteins is markedly better.

What I have found is that some pieces of artwork are imho, using a very professional term here----are stinkers. Watercolors in particular because of the method in which the artists blend the colors to make the artwork. Everything can have a minute impact, including the white of substrate that the artist uses to paint the original on. You may find using a darker neutral background (dark grey or black) better than white as it is possible for lights from your strobes hitting the outside surrounding background (read: brighter white background) effecting the captured image exposure, contrast, etc.

Simply put: watercolors can be stinkers. They definitely are more difficult and take more time in post in the reproduction process.

Ken

 

[Viewmaster]

Hi Craig,

I've shot large repro projects for museums, worked in a commercial photo lab doing art repro full time to printed output, and am an artist making the same sorts of things you are trying to reproduce (drawings, watercolors, acrylics, oils). So this may not be of much help other than consolation.

Everyone here has made some good suggestions and they are all worth following up if you have the time and energy.
What I learned from the guy who had been doing art repro full time for 20 years when I started at the lab was that all the tools we have now are still at best an approximation of what you are seeing when you look at something, and reproducing artwork in particular seems to bring out the deficiencies in our technology.

We might be able to get 95% there, or closer! But at the end of the day, sometimes there will be some colors or some aspect of the art piece that the camera just cannot quite capture, and it's not something that will have a gear/tech based solution for. you will have to go in and potentially spend a good amount of time hand tweaking a color or some other aspect to try and bring it the rest of the way home. And even then, you still might not be able to get an exact match. And this is assuming your setup/equipment chain is in good order.

I say this to try and save you some frustration if you research and end up exhausting some of the other good suggestions above and it's still not coming out quite right. This might be the "art" part of the art and science Doug mentioned.

 

[Craig Stocks]

I think I'm ready to admit that there isn't a technical solution for this particular watercolor painting, at least within the limits of my available equipment. I went back and tried various lighting ratios, with and without UV and polarizing filters on the lights, using flat art and standard profiles, and even using a "good" LED light. All resulted in under-saturated colors on the watercolor. g

The best result for the ColorChecker is using the PhaseOne flat art profile giving a delta-e of 1.7. The painting, unfortunately doesn't match well.

Switching to the standard flash profile the delta-e moves to 2.1 but the painting match is closer.

I haven't found any combination where the painting doesn't require tweaking to fully bring out the colors, especially the yellows.

 

[RLB]

I guess I'm not sure why you took the time to post a reply just to say that you can't help. Are you suggesting it's so difficult that there's no hope of ever learning and I should simply give up?

What I'm stating is that there are many variables, far too many to discuss this way and IMHO acheive a substantial positive forward movement. It also depends how exact one is trying to get...the last 15% is the steepest part of the curve.

The only other point to my post is that often one assumes that digitizing original art should not be that difficult, and superficially it appears it would not, but the reality is to acheive a truly accurate result that we do (exceedes FADGI 4stars) is actually quite expensive and challenging. That's what I assumed you were asking...not what can do to get a better result that woudl be "suitable" or "acceptable" for web viewing or small prints.

R