[sm3600-devel] Gain correction, ideas

[Marian E. <eic...@co...>]

Hi Glenn!
Hi *!

There a *very* good news for us all.

I played around with the offset and gain correction tables, and they
behave better as everything I could imagine:

1) The offset and gain correction table values are in fact correlated to
every individual pixel on the scan line.
   Thus, You can "fix" every single pixel individually. 

2) Because the correction is done on the 12-bit scan data, it is
"lossless" and linear(!).

3) The mapping is done, *before* the window and "resolution" is applied.

This means in fact, that the correction data may be *fixed* one and "for
all" time. it needs not to be changed, when the resolution changes.
There is no need to modify the scanning routine. It is now a simple
scanner setup task.

With the gray mode, there is an experimental correction (in
sm3600-gray.c). It adds a static linear swing of 0x600 to the gains.
With a white sheet of paper, this provides for a nearly constant
luminance level.

The remaining question now is, how a robust mapping from the calibration
strip scans to a gain (and perhaps offset) tables can be done.

Glenn, Antonio, how does Your scanner behave, when You to a "make
testscan" with activated

	#define GAIN_CORRECTION

Perhaps we can make it default, as long as we have no real calibration
at work. This could become part of the backends 1.0.5 if we are quick
:*)


-- 
               Mit freundlichen Gruessen / Yours sincerely
    Dipl.-Ing, Dipl-Theol. Marian Eichholz, Koeln, Germany
                                     pos...@fr...
                              mailto:eic...@co...
                         http://privat.schlund.de/eichholz