I have an I1Studio and I know full well there is nothing I can do to
make it something else and I wouldn't even try, I know it's a waste of
time. I was thinking more of as you said, just improving the
reliability and repeatability and of course the confidence interval as
one average more samples. I recall number like 30 and 40 samples, maybe
90 and 95% C.I. I'm not sure but that many sample may not be necessary
anyway.
My main goal is to be able to make an as excellent profile as possible
with my instrument for maybe a dozen papers at least for now. I'm an old
guy, I speak french mostly, my memory is like Swiss cheese so sometime I
have a hard time understanding your explanations in the docs. I'm also
making conscious effort to minimize waste as much as possible.
1. If possible and useful I'd like to first verify an existing profile,
I just hope it can be done with minimum waste. Something like was
suggested before by Alan maybe sufficient, I don't know, is it?
2. I'd like to learn more on what constitute an excellent profile and
if possible improve on existing one as was also suggested by Alan,
again maybe it's both a waste of time and paper, I don't know, is it?
3. May be I should just forget all the above and find an intelligent
way of selecting an optimal number of patches with the goal of
creating an excellent profile. In the long run it may be the least
wasteful approach. I say this because I use GamutVision to visualize
the gamut of some profile and more and I often see "hole" in the
gamut and other bad behavior and I'm not convinced such profile can
be improved, again, I don't know.
If someone can explain how the I1Studio software can produce a useful
profile with so few patches? I know they offer this optimize thing based
on image color to "improve" the profile and after a while and a good
sample of various images one may end up with an excellent profile. So
why not start from a large enough set of patches and make an excellent
profile from the start? They want us to think this approach is less
wasteful while in fact it takes more sample and more paper in the end.
I don't pretend to be an expert on the subject of statistical
distribution and I managed to figure out the even distribution was what
I call a uniform distribution but the others ones.
_Optimized Farthest Point Sampling_
_Incremental Far Point Distribution algorithm_, I think the key word
here is algorithm, so it's not a statistical distribution per say, just
a mean to spread the test point as far as possible from each other, right?.
_a quasi-random, space filling distribution both in perceptual and
device space_
_a body centered cubic distribution __both in perceptual and device space_
Would I be right in thinking the other "distribution" are also more a
kind of "algorithm" as well? Is there any documentation on those
somewhere I can read so I can understand what they do and how?
You also mention the full spread test patch algorithm, would this the
brute force approach I was talking about above?
Would there be better suited algorithm to use with an existing profile
and how to determine the number of total patches and or the number of
neutrals is there any guide lines for this?
For example, I have a Canon Pro 100, only 8 inks, so would it be helpful
to add a certain number of patches for each ink?
All this is very nice but the printer driver have a roll in all this as
well, I guess one wouldn't use a setting for a glossy paper when trying
to make a profile for mate paper and vice-versa. It's obvious it would
require a lot of paper to find the optimal setting for each paper. One
of the reason to start from an existing profile but again it doesn't
mean they went through all the possible setting. I ask all these
question mainly because I don't want to spent a fortune on papers and
inks and through most if not all of this to waste.
Another problem I observed for some paper profile, they have only one
rendering intent that seem to do something useful. So in such case,
would it be possible to fix and improve such profile?
Sorry for all the question.
Thanks,
YG
On 12/16/2018 5:30 PM, Graeme Gill wrote:
Yves Gauvreau wrote:
Hi,
Since my instrument is not top of the line (I1Studio) I thought of makingthe primary differences between something like the i1Pro and the i1Studio is not
multiple reading
and use Argyll "average" to remove outliers as much as possible before further
processing,
any suggestion on this as well, I like the median approach but which one?
the reliability or repeatablity of the measurement process.
(In fact the i1Studio has higher repeatablility than the i1Pro, due to the use
of a more stable LED light source.) So making multiple measurements and
combining
them is not likely to be something that somehow makes the i1Studio more like an
i1Pro.
You may find benefit in doing this, but it would be to address the
repeatability and reliability of the measurement process, not something
that will have any influence on absolute accuracy.
Cheers,
Graeme Gill.