Making a composite profile, combining ti3 & colprof -r question

[Derin Korman]

Hello,
For experimenting, I combined two ti3 files (of two different hutchcolor
targets), simply grabbing the patch data from one, adding a prefix of a
letter to patch namez (so V1 becomez ZV1 for example) and pasting them on
the bottom of the others, and updating patch count to match. This yielded
much greater delta E but perceptually better profiles I think. I was
wondering if there is any reason I should not do this. What I was trying to
overcome is channels crossing over each other near highlight/shadows and
causing casts/noise

Secondly, I tried employing the -r parameter with colprof, but trying any
value from 1 to 30, I could not get the delta e to change even one decimal
point, does this imply that it does not have any effect? I also tried the
-V but that also didn't change delta E (should it?). When I enter -r 0,
that changes things, but I don't own a device that has 0.0 percent error :)

here are the two ti3 files and the frankenfile if you'd like to take a look.


https://www.dropbox.com/s/o6l2qi1uqc1tsty/timTangoFranken.ti3

https://www.dropbox.com/s/dj8ekprg7y4c6jb/timFuji6830_Jun2014.ti3
https://www.dropbox.com/s/jy2j8cw95by7ew1/timKodak6821_Jun2014.ti3

cheers

*~d*

[Graeme Gill]

Post by Derin Korman
For experimenting, I combined two ti3 files (of two different hutchcolor
targets), simply grabbing the patch data from one, adding a prefix of a
letter to patch namez (so V1 becomez ZV1 for example) and pasting them on
the bottom of the others, and updating patch count to match.

Hi,
there is no need to change the patch names, just to get
colprof working. They are just used for diagnostics once the
charts have been read.

Post by Derin Korman
This yielded
much greater delta E but perceptually better profiles I think. I was
wondering if there is any reason I should not do this.

It's perfectly reasonable. Better to use a larger chart in the first place,
if you are after better characterization though.

The high delta E hints that the process (i.e. print stability, printing process
or measurement process) is not entirely stable.

Post by Derin Korman
What I was trying to
overcome is channels crossing over each other near highlight/shadows and
causing casts/noise

Is this CMYK or and "RGB" print path ?

Post by Derin Korman
Secondly, I tried employing the -r parameter with colprof, but trying any
value from 1 to 30, I could not get the delta e to change even one decimal
point, does this imply that it does not have any effect?

The values are too high - you have hit the lookup table limit. Normal
ranges would be something like 0.0001 to 5.0 %.

Post by Derin Korman
I also tried the
-V but that also didn't change delta E (should it?).

No, that's for video display profiles, and works in combination
with the test patch distribution.

Post by Derin Korman
When I enter -r 0,
that changes things, but I don't own a device that has 0.0 percent error :)

See above.

Graeme Gill.

[Derin Korman]

Hello,

Post by Graeme Gill

Post by Derin Korman
What I was trying to
overcome is channels crossing over each other near highlight/shadows and
causing casts/noise

Is this CMYK or and "RGB" print path ?

This is a drum scanner coming in with RGB, going to an rgb working
profile such as AdobeRGB before printed, channel crossing visible in
rgb working space.

Post by Graeme Gill

Post by Derin Korman
The high delta E hints that the process (i.e. print stability, printing process or measurement process) is not entirely stable.

I meant that if I take A.ti3 and make a profile, it will give 1 delta
E, B.ti will give 1 delta E also on it's own, but once I combine the
ti3s it gives a delta E of 2-3. (symbolic values) Is this because of
the change of standard deviations? They are the same ti3 files, so
there is no changes in the process or input values to colprof

thanks a whole bunch,


*~d*

[Graeme Gill]

Post by Derin Korman
This is a drum scanner coming in with RGB, going to an rgb working
profile such as AdobeRGB before printed, channel crossing visible in
rgb working space.

Hi,
I'm afraid I'm not sure what you mean by "channel crossing" in
an RGB space, since there is no extra degree of freedom to cross over in,
like there is in CMYK space.

Post by Derin Korman
I meant that if I take A.ti3 and make a profile, it will give 1 delta
E, B.ti will give 1 delta E also on it's own, but once I combine the
ti3s it gives a delta E of 2-3. (symbolic values) Is this because of
the change of standard deviations? They are the same ti3 files, so
there is no changes in the process or input values to colprof

There is a table lookup that computes a heuristic smoothing factor that
depends on the dimensionality, sample point density, and sample point deviation. By
duplicating the points you've fooled it into thinking there is a higher
sample point density than is actually the case, and it's used a higher smoothing
factor, resulting in a less tight fit.

Self fit doesn't say much about how well the device is being modeled though,
just how consistent the sample points are.

Graeme Gill.