Brad Funkhouser
2014-02-27 22:08:54 UTC
Graeme,
For weeks now I've been experimenting with Argyll every day between customer
projects. It's helped me immensely to advance my understanding of the color
management process.
The most important thing I've learned is that the ColorCheckerDC input
target has always been inadequate for my needs.
Creating standalone light trap black and PTFE white patches helped me hone
my camera RAW processing.
Figuring out that I could fold in custom patch data and/or combine data from
multiple charts into a single .ti3 file yielded tremendous testing
flexibility. Comparing different input targets, each with its native white
patch AND each with inserted standardized PTFE white patch data helped me
see the differences between relative and absolute mappings very clearly.
Treating a camera/profile combination as "a poor man's colorimeter" and
comparing those XYZ results to i1pro measurements was a great learning tool
and helped me understand the shortcomings of some input targets.
Taking i1pro measured reference data out of a .ti3 file, batch feeding it
through icclu into an appropriately sized working space (i.e. no clipping),
then inserting those RGB values back into the original target .ti1 file to
generate a synthetic working space target TIFF of "what the i1pro saw"
helped me evaluate and better understand measurement variability as well as
transformations from camera to working space and working space to printer
space.
Likewise, taking the averaged camera patch device values out of a .ti3 file
and sending those RGB values back through the .ti1 to generate a synthetic
TIFF of "what the camera saw" was also helpful in comparisons and analysis.
Being able to combine camera scanins of multi-page targets into a single
.ti3 file let me use 2500+ patch input sets without altering my normal
camera imaging area. This was a significant breakthrough.
Generating a 600 patch preliminary far point target spread and using that
profile to build a 2500+ patch cubic perceptual uniform spread helped me
visualize the importance of patch coverage. Those cubic uniform perceptual
charts (-I) are so beautiful and mesmerizing to look at that I just couldn't
force myself to randomize them. I paid the price when scanin had trouble
recognizing the layout, but with a little work it turned out okay.
Experimenting with a closed loop environment where my Epson 9900 inks on a
non-oba paper served as both a well defined input space AND a matched
destination space let me see how the building of an input profile differed
from that of an output profile given the same target data, let me see the
effects of going through an intermediate working space versus going straight
from camera space to printer space, and taught me how finicky the whole
white point situation can be (and that it's not such a big deal to clean
white up at various stages in the process).
Turns out that the pigments in my 9900 inkset must be a fairly good spectral
response proxy for those in the pastels one of my artists uses, because
mapping from camera space through BetaRGB and into printer space just
yielded a perfect looking match print. First try. No mods. Amazing. Of
course it's a simple case, with the image gamut completely inside the camera
target interpolation space and the printer space. But still, I love it.
Experimenting and learning will continue. Just wanted to say thanks for
building such a powerful, flexible, verbose toolset.
- Brad
For weeks now I've been experimenting with Argyll every day between customer
projects. It's helped me immensely to advance my understanding of the color
management process.
The most important thing I've learned is that the ColorCheckerDC input
target has always been inadequate for my needs.
Creating standalone light trap black and PTFE white patches helped me hone
my camera RAW processing.
Figuring out that I could fold in custom patch data and/or combine data from
multiple charts into a single .ti3 file yielded tremendous testing
flexibility. Comparing different input targets, each with its native white
patch AND each with inserted standardized PTFE white patch data helped me
see the differences between relative and absolute mappings very clearly.
Treating a camera/profile combination as "a poor man's colorimeter" and
comparing those XYZ results to i1pro measurements was a great learning tool
and helped me understand the shortcomings of some input targets.
Taking i1pro measured reference data out of a .ti3 file, batch feeding it
through icclu into an appropriately sized working space (i.e. no clipping),
then inserting those RGB values back into the original target .ti1 file to
generate a synthetic working space target TIFF of "what the i1pro saw"
helped me evaluate and better understand measurement variability as well as
transformations from camera to working space and working space to printer
space.
Likewise, taking the averaged camera patch device values out of a .ti3 file
and sending those RGB values back through the .ti1 to generate a synthetic
TIFF of "what the camera saw" was also helpful in comparisons and analysis.
Being able to combine camera scanins of multi-page targets into a single
.ti3 file let me use 2500+ patch input sets without altering my normal
camera imaging area. This was a significant breakthrough.
Generating a 600 patch preliminary far point target spread and using that
profile to build a 2500+ patch cubic perceptual uniform spread helped me
visualize the importance of patch coverage. Those cubic uniform perceptual
charts (-I) are so beautiful and mesmerizing to look at that I just couldn't
force myself to randomize them. I paid the price when scanin had trouble
recognizing the layout, but with a little work it turned out okay.
Experimenting with a closed loop environment where my Epson 9900 inks on a
non-oba paper served as both a well defined input space AND a matched
destination space let me see how the building of an input profile differed
from that of an output profile given the same target data, let me see the
effects of going through an intermediate working space versus going straight
from camera space to printer space, and taught me how finicky the whole
white point situation can be (and that it's not such a big deal to clean
white up at various stages in the process).
Turns out that the pigments in my 9900 inkset must be a fairly good spectral
response proxy for those in the pastels one of my artists uses, because
mapping from camera space through BetaRGB and into printer space just
yielded a perfect looking match print. First try. No mods. Amazing. Of
course it's a simple case, with the image gamut completely inside the camera
target interpolation space and the printer space. But still, I love it.
Experimenting and learning will continue. Just wanted to say thanks for
building such a powerful, flexible, verbose toolset.
- Brad