Ben, for years you have written about home brew CC. I have never tried it since I would not know what color to purchase….oil, or watercolor, or pastels. I would not know how this home brew chart could be read. Or what to read it with. Cheers. David David B. Miller, Pharm. D. 3809 Alabama Street Bellingham, Washington, 98226-4585 360 739 2826

Thank you Ben. Chart would be only for my personal use……..Now, if I missed it, what do you do today?

On Sep 18, 2015, at 2:17 PM, Ben Goren <ben@trumpetpower.com> wrote:

what I do today....

David B. Miller, Pharm. D. 3809 Alabama Street Bellingham, Washington, 98226-4585 360 739 2826

Dr. Miller:

<snipped> I might just see where some of my issues come in. Could it be the cardboard x-rite CC? (I refuse to use the passport CC).

The ColorChecker Passport uses the same patches on its ColorChecker Classic page as the cardboard ColorChecker Classic so there is no reason not to use the ColorChecker Passport. There are added benefits on the facing Photo Enhancement page with finer spaced dark grays and light grays and the higher saturated spectrum colors. These extra patches can be used to evaluate your images and profiles. More information on the ColorChecker Passport may be found at http://rmimaging.com/information/ColorChecker_Passport_Technical_Report.pdf.

Workflow is camera white balanced with what is supposed to be a white target on one side, and black, gray, and white on the other side. These three shades are for me to determine, exposure. Then, I use the white only side to determine white balance in camera.

This is one source of your color error. If you examine the spectra of the white patch on this chart you will find that it is not white but very pale yellow. The color is specified as Munsell N9.5 and an average of 12 ColorCheckers of various ages produces L*a*b* values of L* 96.4, a* -1.0, b* 3.1. We perceive the color as white because it is the whitest object in our field of view and the human vision system adapts to make it appear white. Because the white patch is not white, the white patch should be used for setting the exposure only. Instead, you should set the neutral balance, white balance is a misnomer, with the gray patch on this chart. The gray patch is specified as Munsell N5 which I have measured and averaged from 12 ColorCheckers as L* 50.9, -0.4, 0.1. This is definitely more neutral than the white patch and a better choice for neutral balancing. It is also light enough to produce a good signal-to-noise ratio in the camera’s sensor. The black patch is specified as Munsell N2 and should be used to check for veiling glare. Veiling glare can be controlled by using a lens hood and surrounding the subject with a black background. <snipped> Mr. Goren:

<snipped>

...

Workflow is camera white balanced with what is supposed to be a white target on one side, and black, gray, and white on the other side.

While some white balance targets are more spectrally flat than others, nothing, not even Spectralon, is perfectly flat. Click-to-white-balance relies on a nonexistent physical property.

While PTFE references such as Spectralon® and Fluorilon® do not exhibit perfectly flat spectral reflectances, they are the best materials presently available for white reflectance standards and are used as such for calibrating spectrometers by many companies and NIST. PTFE is the only white material I recommend for neutral balancing cameras. The majority of white objects, including the white patch on ColorChecker charts, use a titanium white pigment which is actually a pale yellow, as noted above. If these white objects are used for neutral balancing a camera the resulting images will have a slight blue color cast. Robin Myers

...

While PTFE references such as Spectralon® and Fluorilon® do not exhibit perfectly flat spectral reflectances, they are the best materials presently available for white reflectance standards and are used as such for calibrating spectrometers by many companies and NIST.

PTFE makes for a great reference standard, but, best I know, outside of photographic click-to-balance workflows, it is not used in a manner that assumes that it is pure white.

If you measure the calibration tile included with an i1 Pro -- the one that you have to put the instrument on before every session -- you'll see that it's decidedly less white than PTFE. It's got a curved spectral reflectivity significantly less than 100%. But it's a beautifully smooth spectrum, and presumably the material is very stable. It's plenty reflective enough that it presents an high signal-to-noise ratio for the electronics. PTFE is also stable and has an even higher reflectivity, making for marginally better absolute properties. But it's less durable than the ceramic used for the i1, and I'm sure much more expensive, which would explain why X-Rite used the ceramic instead.

The calibration plaque for the i1Pro, and its predecessors (e.g. Spectrolino, SPM, ColorEye) use a white material which has been mixed with a binder, I suspect an epoxy, it is not a ceramic. If you closely examine the underside of an i1Pro calibration plaque (not the i1Pro2 because the plaque bottom is hidden) you will see that it has been formed while liquid into the plastic holder, allowed to harden, then the top side has been smoothed and polished.

You don't need an absolute white reference for spectrometry. The instrument (or software) knows what the combined reflected spectrum of its lamp and reference is supposed to be; whatever it measures at the time of calibration is used to create an offset to correct for whatever momentary conditions from the environment or whatever are causing drift. You could use a lump of coal or a colored ink or whatever for that, save the readings would get noisy.

<blockquote>PTFE is the only white material I recommend for neutral balancing cameras.</blockquote>

If you're doing the click-to-balance method, PTFE is as good as it gets. Thread seal tape (like what plumbers use) is a good, cheap source; layer it up to ensure opacity. BabelColor used to sell an affordable 1" circular Spectralon (or equivalent) target, but that was years ago. Actual Spectralon targets are available for insane prices.

Plumber’s tape is skived from a solid PTFE rod into a very thin layer. While it has a very uniform reflectance, as you note, it is very thin and thus translucent. To use it for a white reference requires making a stack of many layers. This makes it difficult to put into a re-usable form. Taping the layers does not work well since PTFE is Teflon®, notorious for being non-stick. I found one adhesive which worked with the PTFE tape, but the adhesive had a slight violet color, which showed through the PTFE tape and affected the color. Mechanically binding the tape layers into a holder is one way I was able to use the tape but it was inconvenient.

But...there're two cheap alternatives that every photographer should be aware of.

Thanks for jogging my memory about alternatives.

First is Tyvek. It's nearly as good as PTFE, and you can buy it cheap, usually in the form of envelopes, at your local office supply store. It's nearly indestructible, which is why it's used for envelopes. It's got a bit of specularity to it, including grain in the specularity, so you have to be careful with lighting. At the same time, if it throws no specular highlights, neither will whatever you photograph.

Tyvek® is an alternative with some caveats. It is available in two forms, Type 10 and Type 14, that might be useful to a photographer. Type 10 is a hard, paper like, stiff material, used, as you mentioned, for mailing envelopes, DVD sleeves, building moisture wraps, etc. It is created by heating and pressing fibers of high-density polyethylene (HDPE). It has a very high reflectance. However, because it has a highly textured surface the material varies in brightness across the sheet. It is also translucent, so it requires stacking many sheets to prevent showing the backing. Further, the texture makes any point sampled white balance vary, depending on the density of the material. If Tyvek is used for white balancing, I recommend averaging the reflectance across a large area. Depending on your camera and workflow this may, or may not, be possible. There is another version of Type 10 Tyvek. It has been produced in a form suitable for printing in inkjet printers. It is used as a water-proof paper for maps, drawings, labels and other purposes. It has a very smooth matte finish and is more opaque than the envelope Tyvek. Unfortunately, it is loaded with fluorescent whiteners, so it is very bluish and cannot be used for a white reference. Type 14 Tyvek is a cloth-like material which is soft and flexible. I have one sample of this material but it was crumpled before I received it so my testing has been hampered by all the wrinkles. It also appeared to have variable density across the material, so I need further samples before I can recommend it for neutral balancing.

Second...is styrofoam. I'm not aware of any commercially-available white balance target other than Spectralon that's as good as a styrofoam coffee cup. As a bonus, the conical shape lets you get directional samples in mixed lighting. Or a styrofoam sphere from a crafts store could do the same.

Styrofoam is a very good white, but it also has the issues of Tyvek in relation to translucency and texture. It is soft so easily damaged, but I have used it in various imaging situations where the foam would not be subjected to handling. Styrofoam cups can also pass their contents into the cell structure, thus coloring the foam, so be sure to get an unused cup. The benefit of styrofoam is very low cost and easy availability. While we are on the subject of cheap neutral balance materials for photography, often an object within the image can be used to balance the image. In some situations I have used clouds as gray references, although this is only with clouds that do not show any coloration, so usually this is for mid-day shoots. Robin Myers

<snipped>

...

To use it for a white reference requires making a stack of many layers.

It's been some time since I last did this, but I know I didn't do anything exotic and still got as good a measurement with an i1 Pro as I do off of Spectralon. Don't quote me; experiment...but I think all I did was fold several layers, wrapping into a square roughly the width of the tape...and I just used Elmer's to glue the bottom layers together and to affix to the chart. I'm sure I must have used tweezers or something similar to do all the folding. I might or might not have wrapped the tape around something like a square piece of the same Canson Platine Fibre Rag I used for the printed patches on the same chart.

I tried various adhesives, all failed to hold the PTFE tape permanently. Some, such as Elmer’s were not used because they fluoresce or they turn yellow when dry. The only one which worked for me was a binary liquid adhesive designed for difficult to adhere plastics, such as HDPE. Although, as I mentioned, it had a noticeably light violet color.

Anyway, it's not that hard to work with, even though the end result is, granted, on the delicate side.

I think, today...I'd probably get some super-wide (1" or wider) tape, thickest I could find, and wrap it around a white-painted block, layer by layer, gluing each layer independently to the back side. When done, I'd probably mount the block in an hole cut to size, or something like that, and use it all by itself without trying to integrate it into a chart. Call it a $1.00 DIY Watch Your White. Fragile but cheap enough to be disposable.

...

There is another version of Type 10 Tyvek. It has been produced in a form suitable for printing in inkjet printers. It is used as a water-proof paper for maps, drawings, labels and other purposes. It has a very smooth matte finish and is more opaque than the envelope Tyvek. Unfortunately, it is loaded with fluorescent whiteners, so it is very bluish and cannot be used for a white reference.

I have a 44" x 50' roll of Canon's official Tyvek Banner Media. The front, printable side is the exact same coating as Canon uses for their Heavyweight Coated paper. The back is the same uncoated textured surface as the envelopes. The back will _not_ take ink; it wicks and smears and makes a miserable mess. But, it's also the same 98%+ flat reflective spectrum as the envelopes. It's a good banner material, and the back side is good for situations where you want a large sheet of spectrally flat white.

Consider yourself very lucky. Canon discontinued this media some time ago. It was one of the materials I tried to get for a white reference product I was developing but only a few isolated samples were available so I could not use it. As part of this white product project I also tried the plumber’s tape, Tyvek paper, styrofoam, various plastics, and many other materials, some quite exotic. Trying to get a material with the right spectral reflectance, good cost, material availability and production abilities is not trivial. I am still looking.

...

Styrofoam is a very good white, but it also has the issues of Tyvek in relation to translucency and texture.

You remind me...two more properties about styrofoam cups. Their reflectivity is in the 80% range, if I remember right, but still flat, flatter than any of the commercial products other than Spectralon. But the translucency of the cups offers another very interesting possibility: you can put the cup over the lens and get an integration of all the light in the scene. Plus, in my experience, the resulting transmission gives a meter reading that's likely right about where you want your exposure to be. Obviously, this can vary from cup to cup, so experimentation and caution would be called for. But, again...that coffee cup outperforms the expensive commercial alternatives....

...

While we are on the subject of cheap neutral balance materials for photography, often an object within the image can be used to balance the image.

This is true...to an extent. Well, it can get you in the ballpark if you've nothing better. But, in practical terms...most white fabrics are a bad idea both because of the fluorescent whiteners in laundry detergent and the translucency letting skin tones through. Papers, especially office papers on the desk or tacked to the wall, are bad because of the optical brighteners and low opacity. White walls are almost always intentionally off-white. White ceilings, even if actually white, tend to reflect the tint of the surroundings.

Ideal is to get a spectrophotometric measurement of the actual illuminant and work from there. But I'm having some luck in some experiments making some educated guesses. In theory, if you know the nature of the illuminant but don't know the exact color temperature, you can nail it down from a picture of a ColorChecker. That's certainly the case for single-source illuminants, such as anything incandescent or daylight; there'll be an idealization of the spectrum that minimizes the errors between predicted and actual chart values, and that'll be very close to the same thing you get from a spectrometer.

I haven't started experimenting with fluorescents, but I'm hoping I'll be able to get to the point where, if I know the scene has fluorescent lighting and the lighting is one of the standards, I'll be able to identify <i>which</i> standard -- again, just by trying them all and seeing which fits best, and if that fit is "close enough.”

I have been involved in both the research and production sides of fine art reproduction for many years. I started with tungsten-halogen lighting, then, because of the poor emission in the short wavelength part of the spectrum, and its emission inefficiency, only about 16-25 lumens per Watt with the rest of the emission being infrared and the potential thermal damage to the artwork, I switched to fluorescent lighting. It is much more efficient, 80-110 lumens per Watt, but it is difficult to use for fine art reproduction due to its diffuse nature. It is, however, very good for product imaging. I use tungsten-halogen lighting only for my infrared imaging projects. When HID lighting appeared on the market I switched to it. With a luminous efficiency of about 80 lumens per Watt, but directional emission, there was more light projected onto the artworks, giving much better exposures and it works very well to make texture apparent in the image. The best fine art reproduction lights I have found are the North Light Products Copy Lights. I use two of their 900 W units. They work well with artworks from business card size to wall size. Although when I assisted in the capture of the Norman collection for the Crocker Museum it was necessary to use 4 North Light units since the Normans were as large as 12 x 9 feet in size.

Mixed lighting should also be possible, but much more difficult since you've got multiple variables.

But...at some point you have to do a sanity check. There's no point trying to do reprographic work in real-world scenes with uncontrolled lighting. You need to abandon that approach at some point and instead either go for reportage, for reproducing the scene as a spectroradiometer might (which is trivial; create your test values with simulated D50 and use the white balance predicted by your simulation for D50 regardless of the actual scene lighting) or instead go "full Adobe" and try for "pleasing" subjective color instead.

There is a fully spectral art reproduction solution available; ColorPony from ColorYoke. It uses a proprietary set of calculations that require spectral measurements of the lighting, spectra of the artwork’s pigments, spectra of a white card used to even out the cosine-fourth and lighting falloffs, plus the spectral response for the camera. The user provides the first three of these items and ColorYoke has the camera spectral response data for several cameras. This result of all this is a version of the artwork image in Adobe 1998 RGB, ready for whatever. It is extremely accurate, more so than any ICC profile I have tested, and I have tested almost all of them. It is not cheap, but for the professional fine art reproduction photographer it pays for itself in a very short time. Robin Myers

On Sep 18, 2015, at 9:34 PM, Robin Myers <robin@rmimaging.com> wrote:

I tried various adhesives, all failed to hold the PTFE tape permanently. Some, such as Elmer’s were not used because they fluoresce or they turn yellow when dry.

I'm remembering more now...I'm sure I wound up wrapping the tape around something opaque and gluing the back side. At that point, even something like pitch would work and not affect the optics. Consider this: use spraymount or the like to securely glue a sheet of Tyvek to some white foamcore. (Top-quality OBA-free fine art inkjet media might be an alternative to Tyvek worth considering. I suppose in theory the foamcore would ideally have an OBA-free paper surface, but, in practice, even black foamcore is likely to work fine, so I wouldn't worry.) Use a razor to cut out a square the same width as the PTFE tape. Cut a single length of PTFE tape about three times its width. Lay the foamcore square with the Tyvek down on top of the tape, in the middle. Wrap one end of the tape to the back, gluing it with whatever ugly gunk you've got that'll hold. Do the same with the other end, only use enough tension to smooth out the surface. Repeat with another piece of tape at 90° to the first, and keep adding layers until you've reached the desired opacity. If you've started with Tyvek at the bottom, it shouldn't take more than a few layers of PTFE. Whether or not you mount the finished reference in some sort of frame is up to you and what you plan on doing with it.

Consider yourself very lucky. Canon discontinued this media some time ago.

Thanks for the heads up! I'll not use any more for banner-type stuff, and save it for the Tyvek.

Trying to get a material with the right spectral reflectance, good cost, material availability and production abilities is not trivial. I am still looking.

It's a good part of the reason why I started looking for alternatives to profiling based off of reflective media. It started just as an attempt to expand the gamut past the relatively low saturation possible from pigments compared with monochromatic light sources...but I eventually realized that depending on a reflective sample for any sort of absolute property that doesn't actually exist is not such a good idea. I still use ColorCheckers for reference, but I could do everything with just a single patch of literally any color that I had a spectral measurement of. Using (and averaging) all 24 patches of a ColorChecker is overkill, but trivial to do...and the surface texture of the ColorChecker's patches is reasonably lambertian, the original cardboard has nice big patches for the studio and the Passport its protective case for the field...they're fantastic products for a color reference; I just don't use them to build profiles, or assume that any of the patches are perfectly neutral, and so on.

I started with tungsten-halogen lighting, then, because of the poor emission in the short wavelength part of the spectrum, and its emission inefficiency, only about 16-25 lumens per Watt with the rest of the emission being infrared and the potential thermal damage to the artwork,

These days, it's worth considering regular dim household incandescent lights, and averaging however many long exposures you need to get the noise under control. Even if the blue channel winds up underexposed by, say, four stops compared to the red...well, just take 2^4 = 16 exposures and perform a median blend on the lot. The blue winds up with the same noise as the red originally had -- and, of course, the red has four stops less noise as well. Or, in other words, by blending 16 exposures, you increase effective ISO by 4 stops, making an ISO 100 exposure equivalent to an ISO 6 exposure. With that sort of approach, you could get superlative results from candlelight...assuming you also did some sort of flat field adjustment since there's no way you're going to get even illumination across the scene from a candle or three.

There is a fully spectral art reproduction solution available; ColorPony from ColorYoke. It uses a proprietary set of calculations that require spectral measurements of the lighting, spectra of the artwork’s pigments, spectra of a white card used to even out the cosine-fourth and lighting falloffs, plus the spectral response for the camera.

I'll bet you a cup of coffee that my approach bests that...if for no other reason than that they're probably not including the spectral efficiency of the lens in the model -- and you wouldn't believe how much variation there is from one lens to another. Nobody does, except in the naive sense that, if you use the same lens to photograph a work as to photograph the chart you use to build a profile, the lens gets "baked into" the profile. But the chart approach falls short on so many other grounds.... What I've got figured out now is likely going to keep me very happy for quite some time. There're three directions I'm thinking of taking it in the future. The first...the camera's spectral response isn't a perfect match for the standard observer, so there're some metameric mismatches in both directions -- colors humans can distinguish but the camera can't, and colors the camera can distinguish but humans can't. In theory, there should be some way to use colored filters to make the camera better simulate the standard observer -- perhaps even a single filter, but more likely multiple filters, and then taking three exposures with each filter and extracting and re-merging the appropriate channels. And, since there's no way I'm going to start designing color filters with specified transmission characteristics, that'll depend in no small part on the luck of the draw and whether or not any readily-available filters are "close enough" to what I'll need. (I could already trivially generate a graph of the ideal filter, but there's no point until I'm ready to tackle the project.) Second, Dr. Berns and his crew at RIT have published lots of papers on true multi-spectral imaging, including approaches that accurately reconstruct spectra using multiple filtered exposures from a digital camera. It might be worthwhile to implement some of that research...but only if I move past the proximate goal of making a print on my own large-format inkjet. The third would be designing an actual spectral-based capture workflow. I already have some ideas for how to make an oversized spectroscope that could use a scanning motion to image something with a DSLR at arbitrary spectral resolution, as fine as 1nm or smaller. But the postprocessing would be a real nightmare, and the mechanics of the setup...and, again, it would be gross overkill for general reproduction work and only of interest to curatorial and preservation work.... Cheers, b&