[argyllcms] Re: testing
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Mon, 15 Oct 2018 11:12:11 +1100
Yves Gauvreau wrote:
Hi,
this as much as possible. As for high bit depth, my thinking is that with
these bigger volume of colors, the distance between to colors should
increase and thus artefact like banding should be worst then with a smaller
distance between colors.
certainly worth keeping the image files in high bit depth when
encoded in a large gamut space.
The reason I'm writing this in the middle of the night for me, is that I
wonder if a similar approach using knowledge about frequencies and visual
perception would make it possible to kind of amplify the signal well above
the noise level to make it easily discernible that this print can only have
been made using a 16 bit printing process or not.
There may be ways of doing this, but beware - the screening process
used by a driver/RIP/printer may or may not be high bit depth, and
screening is a sampling process and therefore can be subject to
many aliasing artifacts. Such artifacts may reveal quantization levels,
but it may be difficult to know where in the process they are.
Some aspects of screening to keep in mind:
* The number of levels the screening process supports, depends
on its details. For fixed screens this can never be more than
the number of dots that make up the screen times the number of
dot sizes/densities (i.e. for a bi-tonal printing process to
support 16 bit, it needs to have 65538 dots, and you need to
average the density over all those dots).
* Screens for the different channels always interfere with each other
(moire) to one degree or another. The printers precision in laying
down the dots can affect the color and gamut, since you get slightly
different ink mixing for dot on dot vs. dot not on dot type screens.
* Multilevel screens (due to printers having different dot sizes
and/or light or light-light inks) have crossover issues. If you simply
stop using one ink or dot size and switch to another, this will
create a visual discontinuity (banding), so the crossover has to
be managed to minimize these artifacts.
* Inkjet printers suffer from banding, due to differences in
nozzle positioning, timing and output volume. Various tricks
are used to cover this up, such as interleaving/weaving,
but these artifacts are always present.
In summary, even if the printing system is high depth thru-out,
you may still see banding like artifacts in the output due to
interactions between the image and the screening and printing process.
The challenge is to somehow come up with a test that reveals
the underlying processing bit depth above all these other
potential visual artifacts.
Cheers,
Graeme Gill.
Other related posts:
