[sane-devel] Fujitsu JPEG Duplex Scanning

[m. allan noah]

On 7/5/08, m. allan noah <kitno455 at gmail.com> wrote:
> On 7/5/08, Simon Glass <sglass at bluewatersys.com> wrote:
>  > Hi,
>  >
>  >  On a Fujitsu FI-5750C I get JPEG working fine on the flatbed, or when
>  >  using ADF on the front / back size simplex.
>  >
>  >  But when I scan duplex JPEG, I get an image which is only half as wide
>  >  as it should be, and seems to have 8 lines of valid data, then 8 lines
>  >  from somewhere else, etc. Also the JPEG apparently has about half its
>  >  size in extraneous data (I get a warning when loaded into gimp).
>
>
> fujitsu duplex scanners buffer the backside in their internal ram, and
>  you can request all of the frontside, then request all of the
>  backside. but, if the machine does not have enough ram to hold the
>  backside, it may error. so, i default to alternating front/back reads,
>  and buffer the backside in the backend. most machines can handle
>  either of those approaches.
>
>  jpeg duplex seems to be a different story- the machines only alternate
>  or serialize, but not both, you have a machine that serializes, so the
>  attempt to descramble it does the opposite.
>
>  there seems to be no way to detect the proper format by querying the
>  scanner, but i might be able to inspect the image data. let me look.

i invented a mechanism to determine the interlacing format while
scanning, and switch modes on the fly. I have tested with a small,
alternating scanner (fi-4120c2) and a large serialized scanner
(fi-4990C). Simon has confirmed that this fixes the issue for his
machines (one alternating, one serial) without hardcoding anything.

i also fixed the sane_get_parameters bug Simon pointed out in another
email. all users of the fujitsu backend should really update to v70,
to get these (and quite a few other) fixes.

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