Open Bug 512705 Opened 15 years ago Updated 2 years ago

qcms gives different results than lcms with a particular profile

Categories

(Core :: Graphics: Color Management, defect)

x86
All
defect

Tracking

()

People

(Reporter: jrmuizel, Unassigned)

Details

Attachments

(2 files)

values like this:

differing output(19): (3 62 204) -> l(66 34 200) vs. q(65 16 200)
Linked image, is too dark, and the ground has to much yellow plus numerous other issues

http://a.imagehost.org/view/0418/forest_stone_2

Jeff, you can set it as multiple OS, im on windows here and i see it on my displays default profile, which i have attached
OS: Mac OS X → All
Summary: qcms gives different results then lcms with a particular profile → qcms gives different results than lcms with a particular profile
(In reply to comment #1)
> Linked image, is too dark, and the ground has to much yellow plus numerous
> other issues
> 
> http://a.imagehost.org/view/0418/forest_stone_2

The image appears identical Firefox* and Safari 4.03 for me. When I extract the image and bring it into Photoshop, it looks the same there as well.

*Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.1.2) Gecko/20090729 Firefox/3.5.2
(In reply to comment #2)
> (In reply to comment #1)
> > Linked image, is too dark, and the ground has to much yellow plus numerous
> > other issues
> > 
> > http://a.imagehost.org/view/0418/forest_stone_2
> 
> The image appears identical Firefox* and Safari 4.03 for me. When I extract the
> image and bring it into Photoshop, it looks the same there as well.
> 
> *Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.1.2)
> Gecko/20090729 Firefox/3.5.2

Yes. The problem afaict is with the display profile and not the image profile.
Confirmed. The errors with these profiles are sloppy. I would close the bug because the profiles are incorrectly built.

The biggest error in the Header tag is the PCS Illuminant is wrong. This may account for the "yellow" effect being reported. And the PCS Illuminant tag should not be ignored by the CMM or corrected for. The display profile is simply bogus.

Also in the Header, the Manufacturer field is invalid. The creator should check the ICC manufacturer registry or leave this blank.

The TRC is defined by 5 points. This is almost certainly wrong, and if it isn't wrong, the display is a piece of crap.
I question the validity of the primaries of the posted profiles as well. I think that's a bigger question than that of the PCS Illuminant. It's more likely that the primaries simply don't represent actual primaries of the display, and also the TRCs likely don't represent the actual TRC of the display either, hence the report that the image is "dark".

In particular, the blue primary is suspect as its hue and saturation is quite a departure from sRGB, Adobe RGB (1998) and any display profile I've got here. As blue is the opposite of yellow, if the profile is describing a more aggressively saturated blue than actually exists for the display, the CMS will reduce much more blue (and bump up more green+red) to compensate, making things look more yellow than they should.

Without measured primaries for this particular display, the above is speculation.
I've got another case of dark image when gfx.color_management.enablev4 is set to true. It looks fine in Gimp for example which use LCMS.

The URI of the image is: https://cdn.arstechnica.net/wp-content/uploads/2016/12/DSC06025-980x551.jpg
Attached file Phillips 258B profile
Attaching my profile, but I guess there are more users affected. See downstream bug: https://bugzilla.redhat.com/show_bug.cgi?id=1403970
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: