Hello,
Just like you wrote. Observed with darktable 1.6.7 on your file. Not
observed with same file using ufraw.
Never observed working with NEF files from Nikon cameras. System is
Xubuntu 14.04 64-bit on i7-2670QM.
Are there the same spots on other photos in the same positions ?
There are black pixels not only in the sky but in the city area at the
bottom of the picture.
For example, on the bottom left find a big triangular gray area. Above
it, a grey building with 3 blue awnings on the right. Near the top
leftmost window of this building there's a black pixel. Near the middle
top of the building facade (actually hiding inner staircase) there's a
cyan cross. These spots make easy to see that a ufraw rendering of the
same sees *nothing remarkable* in those areas. Pixels are just normal.
It's strange that the hot pixel removal tool does not remove at least
the isolated black pixels. They have all the qualities: high contrast,
isolated. When tuned for more sensitivity, that tool *does* find and
remove some (albeit not relevant) "hot" pixels. This should not happen.
Could it be some mathematical error ?
I remember about 15 years ago an open-source DV video decoder did output
extremely wrong color patches in some high-color-saturation areas (e.g.
green patches on a bright orange object). It looked like an overflow error.
To me, there are these possibilities :
* those pixels aren't really black or cyan but have some special value
in the original file that makes them appear so altough they are not seen
so by the hot pixel removal tool (would explain why other decoders don't
see it, yet feels like a flakey explanation)
* these bad pixels don't exist in original file, there is a bug in
darktable's demosaicing code that creates those bad pixels out of some
accident
* these bad pixels exist in original file but there is a bug in the
hotspot removal tool. Perhaps pixels clamped at exact extreme values
aren't processed correctly.
* there is a bug downstream in the processing (which explains why the
hotspot removal tool doesn't find anything remarkable)
IMHO these hypotheses can only be confirmed by analyzing the file more
closely and/or investigating the hotspot removal tool when it works on
this file.
Tried hachoir which doesn't know about this file format.
Is there a way to dump a pre-demosaicing rendering of the image? I know
they're not pretty but this could help.
Regards.
Le 22/07/2015 22:52, Denis Testemale a écrit :
Hello DT users,
during my tests with my new Panasonic LX100, I noticed that I
sometimes can see some "strange" pixels, a bit like hot pixels. Some
are black, some are green. An example is visible here:
http://www.testemale.name/temp/lx100_dt_pixels.png
I don't know if they can be considered defective pixels and I want to
be sure before sending the camera back. What puzzles me is that:
- the "hot pixels" module doesn't do anything
- they are not visible in the OOC jpgs
- they are not visible in Rawtherapee or Corel AftershotPro or UFRAW
interpretations of the same files (only one white bright pixel is also
visible through the other raw converters and removed by DT under some
favorable conditions)
- I can reduce or sometimes remove them by playing with the color
smoothing and treshold slider of the demosaic module, but there is no
reproducible and reliable method that I found.
One raw file is available here:
http://demo2.ovh.net/fr/d19b838bcddddefecadd3b60a757430d/
Can I hear your feedback on that please? Thanks.
Cheers
denis
------------------------------------------------------------------------------
_______________________________________________
Darktable-users mailing list
Darktable-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/darktable-users
--
Stéphane Gourichon
------------------------------------------------------------------------------
_______________________________________________
Darktable-users mailing list
Darktable-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/darktable-users
