[sane-devel] Film scanning and infrared image cleaning

[Julian Hauser]

Hello Klaus,

On Fri, 11.08.17 09:05, Klaus Kaempf wrote:
> Hi Julian,

> * Julian Hauser <julian at julianhauser.com> [Aug 10. 2017 22:58]:
> > Hello everyone,

> > I'm using the Reflecta 10T scanner using the pieusb backend. It mostly works very well, thanks! One thing I have noticed though is that if I enable the infrared scratch and dust removal option "Clean image" in Xsane, the detected defects get painted over with a blurred dot. On film negatives, which have a grainy structure, the results are clearly visible. I've uploaded an example here: https://mis.julianhauser.com/index.php/s/MhSuFltcm0CLIMs. It would be very nice if the corrections repaired the result taking into account the structure of the neighbouring areas (as for example the "healing tool" in GIMP does).

>
> > Does the "overpainting" happen in the firmware of the scanner or is
> > that part of sane?
>
> it's part of sane, resp. the pieusb backend. See
> https://anonscm.debian.org/cgit/sane/sane-backends.git/tree/backend/pieusb_specific.c#n1321
> where various sane_ir_* functions are called.

I've had a look at the functions that are responsible for the dust removal but I have to admit I don't even understand the basic approach taken. Is the author of those functions on the list? It would be nice if someone could explain me roughly what is happening and where a change would have to be made.

I've also noticed that the scratch and dust removal seems to be a bit overeager. If you look at the example image I linked to in the first email you can see that the scan is littered with blurred circles. I scanned some photos with scratch removal on and then off, and it is clear that most of the corrections aren't of anything visible to the naked eye.

>
>
> > Two more things I've noticed with the scanner is that images always
> > come out reversed left-to-right
>
> All images (color, greyscale, b&w) ? In all resolutions ?

All the ones I could test are reversed (but that is only various resolutions in colour mode currently). Given that the same happened with the Reflecta CrystalScan 7200 that I had earlier, I would say it's a general issue.

>
> > and greyscale scans make xsane crash. These things don't bother me much
> > as they are easily fixed with a script after scanning. I can send crash
> > logs if someone is interested in debugging this however.
>
> Please do ! ;-)

Here's the log of an attempt to scan with xsane using greyscale: https://pastebin.com/upMTt9HJ. I've realised that lineart and halftone do not work either.

Thanks,
Julian