Page 2 of 2

PostPosted: Mon Apr 10, 2006 11:12 am
by LSI_Muenier
Dear tobbie,

I am very sorry that I didn't get to read your postings earlier.

I just performed an IT8 calibration with SilverFast on my sample DigitDia4000 scanner. I used a target from our batch production together with the usual batch average reference file.

As a standard checking procedure (which I perform every once in a while) of the generated ICC profile, I measured the resulting Lab values of the scanned target and found:
- an average delta-E of 1.1
which is very good, I think.

Comparing with the reference data manually, I found that the colors in the calibrated scan show the correct saturation, none of them were oversaturated. The IT8 target is correctly reproduced by the scanner.

Please take into account that the calibration process does not try to exactly fit each and every color patch, but allows for correction of errors of measurement (caused by dust etc. on the target). In this way, a smooth ICC profile is generated, which is not disturbed by small errors of measurement. On the other hand, there are some very limited instances in which patch colors may not fit into my destination color space (which is AdobeRGB as is default in SilverFast), leading to RGB clamping. For all these reasons I got a resulting maximum delta-E of 7.9.

A remark to an earlier one of your ideas (maybe I misunderstood it):
I would not propose to generate an IT8 target using a calibrated printer. The spectral characteristics of the dyes used for the print will not be equal enough to the dyes of the originals you want to scan. Ideally, you need an IT8 target made on the material you are going to scan. Therefore we offer targets on Fuji-material, too. Generally speaking, the Lab reference data (although commonly used) are only a compromise when compared to spectral reference data. But for almost all users' purposes, working with Lab data is okay.

Best regards
Martin
Developer at LSI

PostPosted: Mon Apr 10, 2006 12:00 pm
by tobbbie
Hello Martin,

...thanks for your reply to my IT8 learning curve documentation thread.
I know now a lot more than when I wrote it initially, the final conclusion
in my previous post.

As I understand now, the simplification to just "RGB" Data from the mixed
colors in the IT8 patches is not independent of the film material.

Which tools are you using to mesure the delta-E?

I have now bought an "all you ever need" set from Wolf F A U S T
(why do you bl*** this name automatically?) and could discover the
many subtle differences a calibration can make.
He was also especially helpful in discussing the many questions I had and
directing me to further sources of information, e.g. www.argyllcms.com

Finally the IT8 target that I have from LSI is OK now and if this would
have been so right from the start, the whole thread was never written.

bye
tobbbie
:-)

P.S.: There are still some other quirks with SF that I have several tickets
attached to. I hope that finally they will be resolved as well.

PostPosted: Mon Apr 10, 2006 1:02 pm
by LSI_Muenier
Dear tobbie,

the tool is an internal-LSI one that I generated during development.
I am happy that your target works okay now.

Best regards
Martin