[argyllcms] Re: Homebrew ColorChecker

On 2010 Feb 13, at 7:38 AM, Pascal de Bruijn wrote:

> Even  while being  advised against  it, I  still went  ahead and
> tried to make my own ColorChecker:

Since you're printing  this on an inkjet, there's  no advantage to
using the same colors as on the MacBeth ColorChecker. You might as
well use targen to create your  own set of patches -- and probably
a (much) larger set, to boot.

I've been absolutely  slammed the past week, and  this coming week
is looking  to be  no better. But  I did  get a  chance to  snag a
bunch  of paint  chips  from  HomeDepot and  I've  compared a  few
measurements against a  ColorChecker. Perhaps not too surprisingly
in retrospect, chips with a  similar visual appearance have a very
similar spectral response. I'd  be willing to bet a  beer that the
automated color matching system at  paint stores these days should
be able  to come  up with at  least a 90%  match, if  not actually
within manufacturer tolerances.

What's  more,  there are  some  chips  that  have a  purer  visual
appearance  than  the   corresponding  ColorChecker  matches,  and
``cleaner'' spectra  to boot. The  process yellow is  one example;
the  paint chip  had  a lower  high-frequency  response, a  higher
low-frequency response,  a sharper transition, and  everything was
much flatter (fewer small bumps and wavy lines in the graph).

And so I've come to  the tentative conclusion that there's nothing
especially  magical  about  the ColorChecker,  either  in  pigment
composition  or color  choice,  for the  purposes  of profiling  a
digital camera. Sure, the colors are designed to be representative
of typical scenes,  but the choices were made  looooong before the
days of digital spectral colorimetric imaging.

I'm starting  to work  towards the following  for my  own homebrew
chart:

    * The BabelColor White-Balance  Target arrived in the mail
      the other day. It'll be on the chart.

    * I've  already  mentioned that  I'll be  making a  simple
      black trap that I'll be including.

    * Once  I get a chance  to inhale, I'll be  getting advice
      from a local art instructor  on what media / pigments to
      choose. Aside from practical  factors such as durability
      and (lack of) specular  reflections, I'll be looking for
      the following:

        * Ideally, at least one pigment each with a single
          tight spectral peak for  each named color of the
          rainbow. I realize this might not be realistic.

        * One  pigment  each  for  a  reasonable  spectral
          match  for   the  18  colored  squares   on  the
          ColorChecker; this  is to  match the  stated aim
          of  including ``real-world''  colors. (Actually,
          there're patches  on the ColorChecker  with very
          similar  spectral responses  to  each other,  so
          I'll probably reduce the 18  to a subset of just
          a dozen or so.)

        * One  each  of every  pigment she's  had troubles
          with when doing art reproduction in the past.

        * I haven't  had a chance to test it  yet, but, at
          least  in  theory,  I  should  be  able  to  get
          neutrals that are flat enough for these purposes
          out of my Canon  iPF8100. If not, I'll work with
          the artist on that, as well.

    * Once I've got the  pigmented and neutral patches decided
      upon, I'll count  'em up and subtract  that from however
      many patches  I think  I can  fit on  a reasonably-sized
      target. I'll  feed  that  number  to  targen. (Tentative
      guess: a couple dozen pigments,  a dozen neutrals, a few
      dozen from targen.)

    * The  whole thing  gets laid out  in either  Photoshop or
      Illustrator,  and  then  printed  (with  lots  of  empty
      spaces) on Canon Watercolor  Paper on the iPF8100. I cut
      out holes for the white target and the black trap, apply
      the  pigments according  to  the artist's  instructions,
      mount it, and the rest of the usual stuff.

    * Then, of course, comes measurements and tests....

Of  course, I'll  post all  details of  whatever I  come up  with,
including exactly what pigments I used and their measured spectra.

Anyway, if  I were  going to create  a homebrew  ColorChecker, I'd
take a  real ColorChecker to the  paint store. If doing it  on the
cheap, I'd just  get paint chips that seemed  like good candidates
for matches,  and measure  them at  home with the  i1 to  pick the
winners. Then, just use scissors and  tape to assemble them into a
chart. If I had money to  burn (and, for whatever perverse reason,
didn't want  to spend  it on  a real  ColorChecker), I'd  hand the
ColorChecker to the  person behind the counter and ask  for a pint
of each color in flat exterior  paint. (It might be a good idea to
compare the spectra of the various base white paints first....)

You  know, I  just remembered: I  have a  not-too-terribly-distant
uncle who's worked as a chemist  for a paint company. I'll have to
get in touch with him, too....

Cheers,

b&

Other related posts: