I've had a couple of good responses off-list that have been helpful.
One of them pointed out that image viewers can introduce noise when
resizing photos. So it's important to view at 1:1 ratio.
Another responder was not able to see the issues that I described.
To make it easier to see the problem, and to eliminate any resampling
errors from image viewers, I've provided the following four files to
more clearly illustrate the phenomenon.
* http://flic.kr/p/dUj7HP Darkroom - The image is seen at 1:1 ratio
* http://flic.kr/p/dUhtvR jpeg export at 100%, and original dimensions
(for reference)
* http://flic.kr/p/dUj7Wn Same image as viewed in EoG with new dots
highlighted
* http://flic.kr/p/dUpJ2Q Same image as viewed in Geeqie (my primary
viewer) with highlights
I was pointed to an interesting blog post,
http://www.darktable.org/2012/06/upcoming-features-new-interpolation-modes-and-better-resize/.
This was pretty helpful. It gave me additional insight as to controlling
the export.
Let me add a little more information about my export settings:
In core options:
"do high quality resampling during export" - yes
"always use littlecms2 during export" - no
"demosaicing for zoomed out darkroom mode" - "at most ppg (reasonable)"
"pixel interpolator" - lanczos3
In gui options:
"don't use embedded preview jpeg but half-size raw" - no (which is the
default)
Quick question... what does that option control exactly? The manual
mentions that this affects thumbnails. Does this mean only the filmstrip
thumbnails only in lighttable and darkroom?
settings of lighttable's export module:
export to file on disk
8-bit jpeg
quality 88% (that's been my normal setting)
intent = perceptual
profile = sRGB
I'm wondering if that intent setting is part of the problem...
I've had a query as to whether or not my monitor is color calibrated.
It's not clear to me if, or how, the calibration of my monitor would be
related to the dots/artifacts and differences in noise that I'm seeing.
But it's a good question. Some months ago I performed a manual color
calibration using on-line tools such as this,
http://www.lagom.nl/lcd-test/ . My workspace is sRGB.
I hope this note better illustrates the discrepancy that I'm seeing. I'd
like to better understand the cause of it.
Kind regards,
Ken
On 02/12/2013 11:37 AM, Ken March wrote:
Hi Salvador,
Thank you for the very good question!
I exported to jpeg @ 88% compression, AND reduced the size to 2048
width (from 4716). The export is as a local file, (versus some on-line
repository such as picasa, etc.).
So I just now experimented with the export options, including 1) jpeg
compression levels (88% versus 100%), 2) the size (reduced to 2048
width versus full size), and 3) format, (jpeg versus png).
I've posted the results for everyone to see, here:
http://flic.kr/s/aHsjDXTttt . (Note that the PNG files are only
available in the "original" size file download.)
In that same directory is the darktable screenshot, as well.
It appears that there is a discrepancy between the darktable
screenshot, and all of the exported files.
I'm curious to know if I can prevent this discrepancy.
If the answer is no, then is it predictable, so that I may work around it?
Ken
On 02/12/2013 10:17 AM, Salvador Laiz wrote:
Have you taken a look at the jpeg quality export options? Is it at
100% quality?
2013/2/12 Ken March <[email protected]
<mailto:[email protected]>>
Greetings!
Currently I'm editing 28 very noisy sports photos. They were all
shot outdoors under a heavy cloud cover, late in the afternoon
near dusk, at 1600 ISO and a slow lens (F5.6 to F5.0) on a 15
mega-pixel camera, (Canon 500D).
After I edited, and exported 4 of these photos to jpeg, I noticed
differences between the noise in the jpeg files versus what is
seen in darkroom.
I'm new to darktable, so I'm wondering if this might be user error!
For one of these files in question, I have a screen-shot of what
darkroom displays, as well as the jpeg, xmp file, and originating
cr2 file available here,
https://www.dropbox.com/s/2wb2p3zzqokpnqg/denoise-differences.zip
(~22MB). (First time sharing via dropbox, so let me know if I've
screwed it up.)
For reference, the denoise tools that I've implemented are:
1. raw denoise – threshold 0.01, blend mode unbounded, opacity 100%
2. demosaic – amaze, edge threshold 0.003, color smoothing 2
times, match greens full and local avg
3. denoise (non-local means) – patch size 3, strength 100%, luma
50%, chroma 100%, blend mode unbounded
4. sharpen – radius 2, amount 0.5, threshold 0.5, blend mode off
The differences are subtle, but exist. In particular there are
two things I've noticed:
First,the noise seen in the pavement in the foreground and of the
bumper sides around the lower portion of the karts is smoother in
darkroom (as seen in the example screen-shot) than as seen in the
jpeg. The zoom level is 1.5:1, btw.
The second discrepancy is that in the jpeg there a couple of
light dots on the leading kart's bumper in the area in front of
his rear wheel, which are not seen in the darktable screen-shot.
Why is there a discrepancy between what is seen in darktable
versus the exported jpeg?
(I hope the answer is user error! :-) )
It will improve my work flow to understand this discrepancy.
Kind regards,
Ken
------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb
_______________________________________________
Darktable-users mailing list
[email protected]
<mailto:[email protected]>
https://lists.sourceforge.net/lists/listinfo/darktable-users
--
Un saludo,
Salvador Laiz
fulgacian - View my most interesting photos on Flickriver
<http://www.flickriver.com/photos/fulgacian/popular-interesting/>
------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb
_______________________________________________
Darktable-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/darktable-users
------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb
_______________________________________________
Darktable-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/darktable-users
------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb
_______________________________________________
Darktable-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/darktable-users
