On Jul 7, 2013, at 6:50 AM, Iliah Borg <iliah.i.borg@xxxxxxxxx> wrote: > On Jul 7, 2013, at 2:21 AM, Maciej Bryński wrote: > >> 2013/7/7 Iliah Borg <iliah.i.borg@xxxxxxxxx>: >>> What are the values in the reference data file for the patches you used to >>> set white balance? >> >> I tried on this (it has the smallest a* and b* values) >> GS11 14.77 15.32 12.62 46.07 -0.00 0.06 >> 0.16 176.17 0.81 0.81 0.81 0.81 > > I see. However Lab values are not enough to judge the spectral response > flatness, one needs spectral measurements to be sure. On my cards GS11 is far > from spectrally flat. As an example, one would hope that your display that you're reading these words on would have a white that lies on the neutral axis (for D50 or D65 or whatever your chosen white point is), but it's as not-spectrally-flat as it gets: three narrow spikes of red, green, and blue. (PTFE (Teflon) thread tape and Tyvek, on the other hand, both have very flat (visible) spectra and are 98%+ reflective.) This is why I'm a fan of using the profiling mechanism to determine the white point. You could have a chart with nothing even remotely close to spectrally flat or even with neutral tristimulus values, and you'll still get perfect white balance and exposure. > QPcard > http://www.qpcard.com/en_b2c/color-reference-cards/qpcard-203-card.html is > not very expensive, and it allows better camera profiles compared to IT8. I think Maciej might have mentioned that he's got a ColorChecker Passport, which is better than the QPCard (which is, in turn, better than an IT8) for camera profiling. And the latest version of Argyll has reference files for the Passport. Maciej, it's also worth considering making your own chart. Assuming you've got a spectrophotometer, all you need is a bunch of artist's paints and a printer. Get as many different paints as you can. Golden Fluid Acrylics is a good choice if you're buying, or just spend some time in a painter's studio. Plan it all out ahead of time. You'll want at least the base paint by itself and another patch (or more) mixed with white. Other mixtures aren't a bad idea. Figure out how many patches total you'll have, how many painted patches, and generate the difference with Argyll. Lay it all out in Photoshop (or whatever), print it on whatever paper you've got that has no optical brighteners and the largest possible gamut (and glossy is fine, since you need to light it in a way that doesn't throw specular reflections even if it's a matte target), and paint squares by numbers. Measure with the spectrophotometer and you're done. You could spend as much on paint as you would on a ColorChecker Passport and have a chart that far surpasses any you can buy commercially -- and have enough paint left over to make many dozens more. Cheers, b&