RT_sRGB exports look ugly on Flickr etc. using Firefox

[Beep6581]

Originally reported on Google Code with ID 1320

- use RT to generate jpg export from Nikon NEF
- view in GIMP or Irfanview --> looks good
- upload to Flickr --> seems to have extreme 1970s look (like an old paper print where
paper has turned very orange) in Firefox 11.0 --> really unusable
- looks fine in Internet Explorer (seems to use color profile correctly)

How can I solve this? Use a different ICC export profile like "sRGB_IEC61966-2-1_no
black_scaling.icc"?
(of course I can try myself but need your expert opinion how to do it to have same
look in Photoshop, Gimp, Irfanview, Firefox and Internet Explorer).

Other users most likely struggle here as well when using RT's default profile.

Might be related to defects 1166 & 1178.

RawTherapee_WinVista_64_4.0.8.3, Windows 7 64bit

Good (RT export converted with Irfanview):
http://www.flickr.com/photos/globetrotter_rodrigo/6922544180/in/photostream/lightbox/

Bad (direct export from RT using attached profile):
http://i.imgur.com/nXyTi.jpg

Bad (using default profile from RT 4.0.8.3):
http://i.imgur.com/PdjGP.jpg

Raw:
http://min.us/mLJVKhz79

Reported by globetrotter.75d on 2012-04-11 21:27:07

---

• _Attachment: V4_default_MTE.pp3_

[Beep6581]

Tried the same in very old RT 3.0.0. Same result: bad in Firefox, fine everywhere else.

Using RT's default profile:
http://i.imgur.com/Heq5K.jpg

Reported by globetrotter.75d on 2012-04-11 21:44:01

[Beep6581]

Google Chrome 3.0   good
Google Chrome 18.0  good
Opera 11.61         good
Safari 5.1.5        bad

So according to http://en.wikipedia.org/wiki/Browser_wars, at least 35% of users suffer
this problem.

Reported by globetrotter.75d on 2012-04-11 22:29:00

[Beep6581]

RT embeds the color profile in the output file, so any color aware app should be able
to render accurate colors.  We are not responsible for fixing color management issues
in all the other apps in the world.

Reported by ejm.60657 on 2012-04-11 22:42:31

[Beep6581]

Clear, can understand you. 
Still have a hint? How can I avoid the problem using RT? Other output ICC?

Reported by globetrotter.75d on 2012-04-11 22:48:54

[Beep6581]

The best you can do is find an output profile that looks right to you.  But since the
color management chain is broken, for other non-color-managed folks it will only look
right using the same browser on the same hardware.

I thought recent versions of Safari and Firefox were color managed?

Reported by ejm.60657 on 2012-04-11 23:04:12

[Beep6581]

Maybe you could look here first:

https://developer.mozilla.org/En/ICC_color_correction_in_Firefox

Type "about:config" in the URL to access the preferences of Firefox. Default mode is
2: "Color management applied only to tagged images.  (Default in Firefox 3.5.)"

Which version do you use?

By the way, i didn't see any noticeable difference between the three images of comment
#1, displayed in FF11 with default preference values.

Reported by natureh.510 on 2012-04-12 03:02:56

[Beep6581]

You can use Martis check page to see if your browser is good or bad:
http://littlecms2.blogspot.de/2011/06/updated-browser-check-images.html
With IE9 I got full and correct color correction with your images, they look fine.

To resolve for downlevel browsers, simply don't use a profile, but simple "no ICM",
which is automatically interpreted as RGB. And RT is faster with it also.

Reported by oduis@hotmail.com on 2012-04-12 06:04:54

[Beep6581]

According to the link you've provided Oduis, Firefox does support color management (well,
that's what says the images on that page). You only have to enable V4 ICC profiles
first (set "gfx.color_management.enablev4" to true and restart your browser).

Reported by natureh.510 on 2012-04-12 08:06:03

[Beep6581]

I think the problem of globetrotter was that he didn't want to force all user of Flickr
to change their Firefox system configs first before they can see his image ;-)

So I'd still recommend to simply set the output profile to "no ICM", which should work
in all browsers by default.

Reported by oduis@hotmail.com on 2012-04-12 14:51:13

[Beep6581]

I actually never paid attention -- what does RT do if one selects 'no ICM'?  Does it
output colors using RGB values from the working space?  If so, then if a working space
other than sRGB is used, colors will be way off on a typical display.  I would have
thought that using sRGB output would be the safest bet, since most displays are reasonably
close to it.

Reported by ejm.60657 on 2012-04-12 15:04:54

[Beep6581]

I recall that 'no ICM' meant sRGB

Reported by michaelezra000 on 2012-04-12 15:41:38

[Beep6581]

Thank you very much everyone!

Comment #9 is the most important one: 
if an average user uses RT to generate a jpg with default settings and uploads it to
any photo sharing community, with most browsers the images will look ugly. I'd say
that 99.9% of users have no clue about color mgmt, and will just think "well, this
guy has a weird taste, all pictures look like in 1970s."

So, what --> I <-- will do: 
1) I will try some export profiles in RT (eciRGB, ProPhoto, AdobeRGB, no ICM, sRGB)
and check the result in the 5 browsers to see how it looks.
2) change settings in Firefox (v4, mode), so images of OTHER users will look right
on my PC.

What I recommend the --> RT developer team <-- is to rethink about the decision of
using RT_sRGB as standard. Maybe "no ICM" or whatever would be a better standard value?

Btw:
I would bet sRGB_v2 as to be downloaded from http://www.color.org/srgbprofiles.xalter
will be best result to work in all situations. So why not make RT use exactly these?
I also try to find out which profile will look best when printing images thru professional
image printing companies. I assume ProPhoto will be a good base; RT could offer a default_print_professional.pp3
or similar. Just thinking ...

Reported by globetrotter.75d on 2012-04-12 17:16:30

[Beep6581]

+1 for setting NoICM as standard (didn't know that it was changed?)

No ICM does not write any profile to the output file (an circumvents the slow LCMS
by the way). A file with no dedicated profile is generally processes as sRGB by practically
any program these days.

Reported by oduis@hotmail.com on 2012-04-12 20:40:00

[Beep6581]

Well, there are two questions here:

1) If NoICM is selected, what RGB values are written out -- the working space, or sRGB
values, or monitor RGB values?  If the latter, are these sRGB if there is no monitor
profile?

2) If the answer to (1) is sRGB color values, then why is RT_sRGB different?

Reported by ejm.60657 on 2012-04-12 20:57:24

[Beep6581]

Color mgmt really seems to be a big problem with many applications. I've done this:
- update white balance on RAW input image in RT for snow to be really white
- export same image with same settings to several color profiles from RT
- in addition save one of these images from RT as jpg in Irfanview
- get all 10 images on each application and do screenshot, then save as jpg in Irfanview

Results are here:
http://vwlnt7vw4897.minus.com/mfT0FlTSk/1f --> there are 9 screenshots, from browsers,
Irfanview and Gimp, use left and right arrows to navigate

For me this means: 
- in about half of the tools photographers use, ALL images exported from RT4 look bad
- only workaround I could find: use output profile sRGB in RT, then do a save-as in
Irfanview without touching the image --> looks great in most applications but is loosing
image quality due to second conversion jpg to jpg
- question for the RawTherapee developer team is, why none of the RT export profiles
seem to result in correct colors but a simple save-as in Irfanview does (for most viewer
applications)
- do not use Gimp if you want to see right colors, maybe I'll finally get Photoshop

Question: how do Photoshop or Lightroom render the images?
Here are the RT export files: 
- as zip to download: http://i.minus.com/mytTRINT/gallery.zip
- to view one-ny-one: http://vwlnt7vw4897.minus.com/mytTRINT/

Reported by globetrotter.75d on 2012-04-12 21:01:22

[Beep6581]

> 1) If NoICM is selected, what RGB values are written out -- the working space, or
> sRGB values, or monitor RGB values?  If the latter, are these sRGB if there is no
> monitor profile?

If "No ICM" is selected, the colors are converted to sRGB manually by matrix. No LCMS
involved, and no profile embedded into the file. Obviously most viewers are not compatible
with the profile embedded.

Reported by oduis@hotmail.com on 2012-04-12 21:13:12

[Beep6581]

About comment #16: are you really sure? 

(1) If I export a file from RT with "no ICM", Irfanview as well as Gimp still say they
do find a profile in the image:
- Irfanview: "JPEG, RTsRGB - Rawtherapee"
- Gimp: "RT_sRGB gamma sRGB(IEC61966 equivalent)"

(2) If I use the file from (1) and do a save-as in Irfanview, the profile is gone:
- Irfanview: "JPEG, progressive"
- Gimp: "sRGB IEC61966-2.1"

(3) when exporting a jpg from very old RT 3.0.0 there seems to be no profile embedded
in the image when using "no ICM":
- Irfanview: "JPEG"
- Gimp: "sRGB IEC61966-2.1"

So maybe the behavior was changed after v3.0.0 and is now buggy?

Reported by globetrotter.75d on 2012-04-12 21:36:42

[Beep6581]

in my opinion priority of the defect should be raised

Reported by globetrotter.75d on 2012-04-12 21:50:20

[Beep6581]

I'm sure that it should circumvent CMS. But I also see that in exiftool it still writes
a profile, which is wrong. I'll check, and yes, the priority is surely higher.

Reported by oduis@hotmail.com on 2012-04-12 21:55:49

• Labels added: Priority-Medium
• Labels removed: Priority-Low

[Beep6581]

OK, here is your patch. The code embedded the profile though it was never converted
with it (because as I said, it circuments RT).
Now when you select "No ICM" it is really no ICM, and not output profile embedded (according
to exiftool).
Do you compile yourself?

Reported by oduis@hotmail.com on 2012-04-12 22:07:11

• Status changed: PatchSubmitted

---

• Attachment: noICM.patch

[Beep6581]

If not, here is the updated EXE only, 7zipped for Win64:
http://www.visualbakery.com/RawTherapee/Download/RT.7z

Reported by oduis@hotmail.com on 2012-04-12 22:10:24

[Beep6581]

Well well... NoICM = RT_sRGB is what has been wanted and fully tested by jdesmis, and
not using any profile may lead problem in some tools that makes Gamut control, like
Vibrance, i don't know. Could you please get in touch with jdesmis for more information
on what he did?

I know that color management is a plugins in IrfanView; is it enabled too? Also which
monitor profile are you using in RT?

So "-1" for changing the NoICM behaviour as of now for me.

Reported by natureh.510 on 2012-04-13 00:10:58

[Beep6581]

What is RT_sRGB?  Does it differ from sRGB IEC61966-2.1, which is the official sRGB
standard?  If not, why do we need to give it a different name?  It is in the public
domain, as far as I know.

Reported by ejm.60657 on 2012-04-13 01:05:22

[Beep6581]

I think the case is simple: when the user explictly selects "No ICM" he wants not ICM.
He can happily select RT_sRGB and have the profile embedded and the correct conversion
it he want's to.

Reported by oduis@hotmail.com on 2012-04-13 06:01:44

[Beep6581]

... and answering Emil: RTs profile is not the official standard sRGB profile, which
is probably causing some of the headaches. Maybe neuer ICC version that is not widely
understood yet.

Reported by oduis@hotmail.com on 2012-04-13 06:08:02

[Beep6581]

.. answering natureh: If the users select "No ICM" the goal is that it should work with
any program and browser. We cannot say to the users "please install a plugin IrfanView,
because our RT images are a bit special".

Reported by oduis@hotmail.com on 2012-04-13 06:10:40

[Beep6581]

Emil, RT_sRGB is same to sRGB but with a gammaLUT of 4096 points instead of 1024. It
gives a smoother histogram at dark shades.

Look at the bottom of Jacques's page

http://jacques.desmis.perso.neuf.fr/RT/ColorRT2_5.html

Reported by iliasgiarimis on 2012-04-13 07:13:42

[Beep6581]

PS: This morning I pointed Jacques to this issue. If there are no more objections I'll
commit the fix tonight.
Please mention that the patch does not delete the special profile, you can use it normally
if you select it.

Reported by oduis@hotmail.com on 2012-04-13 13:55:14

[Beep6581]

no objection :)

Reported by jdesmis on 2012-04-13 14:21:35

[Beep6581]

Patch tested successfully, so ok for commiting, but it seems that the ouput gamma is
broken, if you e.g. select standard_g1.8 (it looks to be a long time bug, not related
to this patch).

Reported by natureh.510 on 2012-04-13 14:59:34

[Beep6581]

Thanks for testing, committed to DEFAULT.
Does one of you both want to fix the gamma issue (I'm currently working on the external
programs feature)? I'd wait then before pushing out another interims build.

Reported by oduis@hotmail.com on 2012-04-13 15:16:52

• Status changed: FixedPendingConfirmation
• Labels added: 4.0, Component-Quality

[Beep6581]

I'm actually trying to figure out which revision is the culprit, but maybe Jacques know
better where to look for.

Reported by natureh.510 on 2012-04-13 15:32:49

[Beep6581]

Ok, i've isolated the culprit changset: 18678158c6e2 / Michael Ezra / 2012-01-26 / Merge
with 1782f47ae9c9699a4a39036884ad9de4df010981

Starting from there, the output gamma is broken. I'll investigate for the bug tomorrow,
even if i'd appreciate if you could look at it first Jacques (you should find it faster
i guess)

Should i open a new issue?

Reported by natureh.510 on 2012-04-14 00:16:59

[Beep6581]

Jacques have corrected the output gamma problem, so this issue can be closed if you
want.

Reported by natureh.510 on 2012-04-14 17:37:57

[Beep6581]

Build is out. Should be warn of selecting the other profiles in the manual? Something
like "Warning, pretty advanced stuff, but not compatible with all viewers"

Reported by oduis@hotmail.com on 2012-04-14 18:49:04

[Beep6581]

Diplomatic warning is in the manual now (thanks Hombre for pointing it out in the forum).

Reported by oduis@hotmail.com on 2012-04-14 19:27:50

• Status changed: Fixed

[Beep6581]

Thanx guys for your efforts, now I love RT even more. 
Marvelous product from a competent international team!
Looking forward to download version 4.0.8.x to work on my ski trip images ;)

Reported by globetrotter.75d on 2012-04-14 19:49:19

[Beep6581]

Thanks Globetrotter :-)

Reported by oduis@hotmail.com on 2012-04-14 20:43:58