[sane-devel] Scan Correction Program! :)

m. allan noah kitno455 at gmail.com
Mon May 12 11:51:01 UTC 2008


The results might be caused by poor (or no) CIS calibration being done
by the backend. You could try getting a trace of the device in action
with the windows driver, and investigate adding the calibration steps
to the backend.

generally this involves doing several small scans of a white area
under the shell of the scanner, with and without the lamp on. this is
usually pretty easy to identify in the logs. interpreting it is a
different matter... :)

allan

On 5/12/08, Ekkehard Morgenstern <ekkehard at ekkehardmorgenstern.de> wrote:
>
>  It's the ma1509 backend.
>
>  I've already seen that there's a possibility to provide gamma correction
>  tables and such to SANE from within a backend. How does it work?
>
>  I'm not sure whether the poor results of my scanner are the result of
>  hardware aging, or if it's just because the backend seems to pass on the
>  data from the scanner chipset unchanged.
>
>  It's an LED scanner, so there must be a set of photo diodes or photo
>  transistors somewhere. It appears as if the scanner sends the data in
>  raw form, unadapted to the characteristics of the semiconductors.
>
>  Distribution of RGB values across their channels suggest that the data
>  should be scaled or transformed somehow. I spent a whole night this
>  weekend trying to figure out some formula that would solve the problem,
>  but I didn't find a solution.
>
>  Instead, I wrote this program to simply compute the difference between a
>  blank page and the color white. The program generates one line of
>  averaged out scaling data, which is stored in the file "white.shape".
>  I'm not sure whether the data can be made constant. If it were to be
>  included in the backend driver, it has to be fairly constant over all
>  scanners of that type.
>
>  After all, the Windows driver must do the same thing somehow! ;)
>
>
>
>  On Sun, 2008-05-11 at 15:42 -0400, m. allan noah wrote:
>
>  > Ekkehard- another choice might be to do this sort of calibration
>  > inside the driver in sane, but that would require becoming familiar
>  > with the code of the sane backend which drives the scanner....
>  >
>
>
>


-- 
"The truth is an offense, but not a sin"



More information about the sane-devel mailing list