[sane-devel] CIS scanners: to calibrate or not to calibrate :) this is the question (my scanner is mad and lazy :)

Rene Rebe rene at exactcode.de
Sun Jun 29 12:27:39 UTC 2008


Hi,

Pierre Willenbrock wrote:
> Rene Rebe schrieb:
>> litlle girl wrote:
>>> AFE settings looks more stable:
>>>
>>> grep "\[gt68xx\] afe" ./*
>>> ./color-1.log:[gt68xx] afe 0x0f 0x04 0x10 0x05 0x0f 0x04
>>> ./color-2.log:[gt68xx] afe 0x0f 0x04 0x10 0x05 0x0f 0x04
>>> ./color-3.log:[gt68xx] afe 0x0f 0x04 0x10 0x05 0x0f 0x04
>>> ./gray-1.log:[gt68xx] afe 0x0f 0x04 0x10 0x05 0x0f 0x04
>>> ./gray-2.log:[gt68xx] afe 0x0f 0x04 0x0f 0x04 0x0f 0x04
>>> ./gray-3.log:[gt68xx] afe 0x0f 0x04 0x0f 0x04 0x0e 0x03
>>> ./lineart-1.log:[gt68xx] afe 0x0f 0x04 0x0f 0x04 0x0e 0x03
>>> ./lineart-2.log:[gt68xx] afe 0x0f 0x04 0x0f 0x04 0x0e 0x03
>>> ./lineart-3.log:[gt68xx] afe 0x0f 0x04 0x0f 0x04 0x0f 0x04
>>
>> Aside exposure and AFE the backend probably uploads a
>> scaling value for each image sensor element (white / dark
>> shading etc.?) that should be pretty constant.
>  >
>> (And no, I still have not read the genesys backend source
>>   to verify what it's doing, this is just from other
>>   educated experience.)
> 
> I don't know what the gt68xx uses for calibration data,
> but the highly experimental genesys/calibration-cache code stores(in 
> memory only) for each combination of color mode and resolution:
> * exposure time
> * AFE offset and gain
> * per pixel offset and gain(actually just the black/white levels, but 
> the offset and gain are fully defined by those)

Yeah, my point was to write them out to some permanent
storage (filesystem) to cache the values and thus avoid 20-30s(++)
every time a user accesses the scanner.

>> I bet the Windows driver also does not re-calibrate every
>> scan, at least the Mac driver for the Canon Lide 70 does
>> only calibrate once after unpacking.
>>
>> Yours,
>>
> 
> Regards,
>   Pierre

-- 
   René Rebe - ExactCODE GmbH - Europe, Germany, Berlin
   http://exactcode.de | http://t2-project.org | http://rene.rebe.name



More information about the sane-devel mailing list