Different programs' generated scanner ICC profiles are so different

[darlingm]

Using the same TIF and source TXT file, getting wildly different scanner ICC profiles from different programs.  I've tried argyll, basiccolor, scarse, lprof, coca, Profile Maker, Profile Prism, and I think a few others.

How does one evaluate which one is best, and choose which one to use?

I scanned an IT8 printed on reflective Fuji paper sold by Wolf Faust, who seems to have a great reputation.

This image is created by Color Think.  I set all the profiles to wireframe so you could see through them.  The easiest thing to look at is the 2D tracing on the "floor" in the image, because in the 3D space, there's just too much to see what's going on.  (Obviously, something's out of wack with the one that contains a wedge in it in the bottom left.)

(As a total side note, how does this scanner's color space rate against others?)

[Ernst Dinkla]

(As a total side note, how does this scanner's color space rate against others?)

Which scanner?

Most flatbeds are nice for CMY dye based originals, RGB filtering and lamp fitting the dye spectral properties: so for photo, film. A Fuji CMY dye based reflective target does not change that. A Color Checker chart may improve the color fidelity for other originals but sensor and lamp can still be the limiting factor for a wider range of colorants than the CMY photo dyes. Scanners differ, my Epson 3200 does it better with watercolor paintings than my V700. I think Epson targetted the V700 more for photo, film than the older versions of the Perfection scanner range.

On your different profiles; operating temperature of the lamp could be an issue, a cold start scan and the scanner not warming up can make a difference. The Epson V700 warms up before a scan when it has been idle for some time.

Good reading:
http://www.image-engineering.de/index.php?option=com_content&view=article&id=57&Itemid=91
second PDF

--
Met vriendelijke groet, Ernst

Dinkla Grafische Techniek
Quad,piëzografie,giclée
www.pigment-print.com

[Rhossydd]

On your different profiles; operating temperature of the lamp could be an issue, a cold start scan and the scanner not warming up can make a difference. The Epson V700 warms up before a scan when it has been idle for some time.

The OP says that all the profiles are created from the same data, so that isn't an issue.

Yes, it's a concern that there seems to be significant differences, but without details of what settings for both building the profiles and comparing them are available or have been used it's not a lot of use.
The bottom line is how do actual scans compare ?

[Mark D Segal]

I laid out an internal consistency check procedure for testing whether the profile generated with a scanner is at least capable of rendering back the values of the patches in the IT8 target used for creating the profile.You need a spreadsheet, a bit of knowledge about filling the cells with formulas given in the article and the reference file for the IT8 target to do this. You can download the article laying out this procedure here: http://www.luminous-landscape.com/reviews/scanners/plustek-pdf.shtml. I did it only for the grayscale, but it can be extended to test any of the patches. Note this is an internal consistency check only.

[digitaldog]

Using the same TIF and source TXT file, getting wildly different scanner ICC profiles from different programs.  I've tried argyll, basiccolor, scarse, lprof, coca, Profile Maker, Profile Prism, and I think a few others.

How does one evaluate which one is best, and choose which one to use?

Doesn’t surprise me at all! There is a lot of extrapolation and proprietary techniques used to build profiles.

How to evlaute? You try them and pick the one you prefer best.