Daylight white balance values

[Beep6581]

Originally reported on Google Code with ID 1315

The daylight white balance value should be tailored to each ICC/DCP profile. Until such
time, I would like to assemble a list of these temperature/tint values and add them
to the manual. Please help complete this list (Oduis wink wink):

Documented in the white balance section of the manual:
https://docs.google.com/document/d/1DHLb_6xNQsEInxiuU8pz1-sWNinnj09bpBUA4_Vl8w8/edit#heading=h.tbejc8i25xz1

MAKE     MODEL     TEMP  TINT
-------  --------  ----  -----
Canon    EOS 400D  4862  1.03 
Canon    EOS 7D    5770  0.971
Canon    EOS D60   4723  1.237
Nikon    D300      5277  1.07
Nikon    D50       5321  1.18
Olympus  E-1       5118  1.154
Pentax   K10D      5420  1.094

Reported by entertheyoni on 2012-04-08 13:40:14

[Beep6581]

Please review this:
https://docs.google.com/document/d/1DHLb_6xNQsEInxiuU8pz1-sWNinnj09bpBUA4_Vl8w8/edit#heading=h.yib89jckz127

Reported by entertheyoni on 2012-04-08 14:40:03

[Beep6581]

"Daylight" is somewhat a broad definition. We would need to be specific in describing
 a particular kind of daylight when providing exact values for temperature & tint.

Reported by michaelezra000 on 2012-04-08 15:21:41

[Beep6581]

How about "Daylight at noon when the sun is at least 45° above the horizon"?

Is this paragraph correct?
"The daylight white balance temperature values used by your camera and commonly referred
to in literature and on the internet are averages. The precise temperature and tint
combination you should use for daylight photos is specific to the ICC or DCP input
profile auto-matched with your camera by RawTherapee."

Reported by entertheyoni on 2012-04-08 15:39:53

[Beep6581]

It seems to me that this is a pointless exercise.  Preset white balances are just ballpark
approximations; they will never achieve true white balance except by accident.  So
different cameras have different ballpark approximations; do we really want to spend
the effort to keep updating (and cluttering) the documentation/code with this?  I suspect
we have better ways to spend our time (Olli wink wink).

Reported by ejm.60657 on 2012-04-08 15:59:20

[Beep6581]

How is documenting them pointless but implementing them was not?

On the left is using the daylight white balance value from the list above, on the right
is using the "Daylight (Sunny)" preset.

http://i.imgur.com/1g7Rw.jpg

Reported by entertheyoni on 2012-04-08 16:54:43

[Beep6581]

My point illustrated -- the right hand image has a strong magenta cast.  The left hand
one looks a bit warm/green.

Perhaps I'm not up to date on all the changes going on with profiles -- do the new
input profiles change the values for various WB presets in addition to providing the
input profile, according to the individual camera?  If not, where in RT is it coded
that the WB presets change according to the camera model?  Or is this a 'feature' request?

Reported by ejm.60657 on 2012-04-08 17:31:37

[Beep6581]

That would be a feature request. As illustrated, the approximated daylight setting in
RT is visually far from what it should be for my camera. I remember the scene the way
it is on the left "5420 1.094" version, not the right "Daylight (sunny) 5300 1.000"
version.

Reported by entertheyoni on 2012-04-08 18:34:44

[Beep6581]

Isn't this what partial profiles are for?  It would seem to me more efficient, rather
than accumulating a long list of things for RT devs to maintain, to empower users to
create their own presets and store them.  We have that capability for whole profiles.
 One thing that we were thinking about a while ago was allowing a dialog where users
specify a partial profile, which when loaded alters only specified tool settings; I
don't think it ever got implemented, but its use case scenario is exactly the sort
of thing you are asking for here.  ACR has this capability, and it is sorely needed
in RT.

Reported by ejm.60657 on 2012-04-08 19:49:06

[Beep6581]

The partial profile has been recently added in the default branch by Hombre and could
certainly be used to store custom WB presets.
I am a bit surprised why this necessary on per camera basis: aren't the input profiles
supposed to normalize out such camera differences in color reproduction within the
limits of each sensor?

Reported by michaelezra000 on 2012-04-08 20:16:29

[Beep6581]

Sorry, didn’t read this thread till now.
The white balance numbers coming from the camera are unreliable (because according
to Thomas Knoll, they are just hints for the cam makers software and are usually skewed
because of their own processes).
Then the profile illuminant, which leads to the ballpark daylight white balance (5000,
5500…). There are two approaches: Adobe usually has dual illuminants, one fixed recorded
at tungsten light, one fixed at cloudy (6500k). Then they mix the twos for anything
inbetween. Now DCRAW just takes the 6500 values as far as I found it. Then X-Rites
software can do dual illuminant (compatible to Adobe), but also single illuminant (which
I use for our profiles, since we only have one profile). Then X-Rite auto selects the
base Kelvin value for the reference image. I then take the recommended base white balance
by selecting the second whites patch.

I think somewhere in the forum I posted to an article of a guy who measured several
cameras with Lightroom. It seems to be normal that e.g. the daylight white balance
is pretty different between the cameras.
So yes, we could read out the illuminant (it's in the DCP) and maybe change the white
value a bit, but the exact one cannot be derived automatically.
And no, this is not a problem of the profiles. Try it with the DCRAW simple profiles,
will yield the same differences. In fact white balance is usually exactly the same
with profile vs. DCRAW.

Reported by oduis@hotmail.com on 2012-04-09 19:43:41

[Beep6581]

PS: DrSlony, just search Google Issues, the closed ones. I documented every Kelvin setting
there.

Reported by oduis@hotmail.com on 2012-04-09 19:53:27

[Beep6581]

Here is some more on this:
http://www.lightroomforums.net/showthread.php?7925-White-Balance-Discrepancy-RAW-vs.-LR-import
Check the posts of Mark Sirota  (good night)

Reported by oduis@hotmail.com on 2012-04-09 20:40:27

[Beep6581]

I updated it with what I could find, there are still a few values missing. Good night!

Reported by entertheyoni on 2012-04-09 21:32:52

[Beep6581]

Then just tell me which are still missing.

Reported by oduis@hotmail.com on 2012-04-10 05:17:41

[Beep6581]

Canon EOS 5D
Canon PowerShot G10
Nikon D3100
Panasonic DMC-FZ150
Panasonic DMC-FZ35
Panasonic DMC-FZ38
Sony NEX-5N
Sony SLT-A55V

Reported by entertheyoni on 2012-04-10 14:03:24

[Beep6581]

Canon EOS 5D: 4993 / 0.998
Canon PowerShot G10: 4885 / 1.078
Nikon D3100: 5087 / 0.955
Panasonic DMC-FZ150: 5950 / 1.02
Panasonic DMC-FZ35: other name for FZ35
Panasonic DMC-FZ38: 4840 / 1.10
Sony NEX-5N: 5832 / 0.890
Sony SLT-A55V: 5254 / 1.081

Reported by oduis@hotmail.com on 2012-04-10 20:01:41

[Beep6581]

Thank you, updated.

When adding new profiles please add the temp/tint values to the manual, at least until
something makes this obsolete.
http://bit.ly/IhEMHT

Reported by entertheyoni on 2012-04-10 20:45:37

• Status changed: Done

[Beep6581]

Thanks, will add it to the live manual in the future.

Reported by oduis@hotmail.com on 2012-04-10 21:33:34

[Beep6581]

If the correct daylight white balance temperature/tint values depend on the input profile,
that means we can document the exact, correct values. Is there any reason not to do
so?

Copied from the obsolete manual on Google Docs:

ICC/DCP Specific Daylight White Balance Values
The daylight white balance temperature values used by your camera and commonly referred
to in literature and on the internet are averages. The precise temperature and tint
combination you should use for daylight photos is specific to the ICC or DCP input
profile auto-matched to your camera by RawTherapee. This will give you a technically
color-accurate image where whites are white, though aesthetically it may not be what
you want.

MAKE       MODEL         TEMP  TINT
Canon      EOS-1D MKIII  4871  1.113
Canon      EOS 20D       4733  0.969
Canon      EOS 40D       5156  1.049
Canon      EOS 400D      4862  1.030
Canon      EOS 450D      4950  1.050
Canon      EOS 5D        4993  0.998
Canon      EOS 550D      4915  0.916
Canon      EOS 7D        5770  0.971
Canon      EOS D60       4723  1.237
Canon      G12           5821  0.994
Nikon      D200          4936  1.064
Nikon      D300          5277  1.070
Nikon      D3000         5302  1.109
Nikon      D3100         5087  0.955
Nikon      D3S           5100  0.970
Nikon      D50           5321  1.180
Nikon      D5100         5621  0.989
Nikon      D700          5000  1.100
Nikon      D7000         5398  0.986
Nikon      G10           4885  1.078
Olympus    E-1           5118  1.154
Olympus    E-P2          5000  1.060
Panasonic  DMC-G1        5550  1.000
Panasonic  DMC-G3        5800  1.050
Panasonic  DMC-GH1       6280  1.036
Panasonic  DMC-GH2       5680  1.043
Panasonic  DMC-FZ150     5950  1.020
Panasonic  DMC-FZ35/38   4840  1.100
Pentax     K10D          5420  1.094
Pentax     K200D         4465  1.129
Sony       A700          5280  1.076
Sony       A900          5258  1.042
Sony       NEX-5N        5832  0.890
Sony       A55 SLT-A55V  5254  1.081

Reported by entertheyoni on 2014-05-05 08:29:38

[Beep6581]

The WB temp/tint do not depend on the input profile (dcp/icc) but only on the d65 colormatrix
used .. which can come from the dng exif data  or camconst.json or rawimage.cc or dcraw.cc
(not sure if this is the exact priority order).

Here lies something I many times asked but I had no clear answer yet .. Jacques are
you reading ?
Are the related WB calculations assuming that the colormatrix used is for d65 illuminant
and so when the used matrix came from a shot under d50 things go to a wrong direction
?.
If so then for all the models in the above list, the WB temp/tints are based on wrong
data as the default dcraw colormatrix was changed (in rawimage.cc) with the colomatrix
of the dcp/icc profile (which is usually in the 5000-5500 range). There was a reason
for this change in the days of icc profiles because there was the need to "match" the
iic profile with the colormatrix in use. I think now that we use dcp profiles there
is no reason to do this ..

The above values are not the correct values of daylight but the values that came from
WBalancing the reference shot which could be shot 
- in different lighting (usually in the 5000-6000K range but can be even lower than
5000K if it was shot in winter ..).
- color casts coming from the lens used 
- etc

I think a calibrated setup like that of Imaging resource's is more reliable than the
user's shots under "daylight"

Reported by iliasgiarimis on 2014-05-05 11:40:38

[Beep6581]

How are these values derived? By taking a spot white balance from the Colorchecker used
to build the DCP profile as described by oduis above? If this is the case, I don't
think they can be used as a daylight standard because no one knows whether the lighting
conditions actually *were* standard daylight. The profiling software does not require
"perfect" daylight, but works by trying to neutralize any color cast, so you get practically
the same profiles for a wide range of (daylight-like) lighting conditions, even with
flash there is hardly a difference in the resulting profiles (there will of course
be noticeable differences with discontinuous spectrum lighting like flourescent etc.).

Put differently, if you build two profiles, one with 6500K daylight, the other with
5000K daylight, and apply these to the same picture and the same WB, the end result
will be practically indistinguishable, and not 1500K apart.

Reported by stefan.ittner on 2014-05-05 11:50:14

[Beep6581]

Stefan yes, I was surprised to find that when I applied DCPs made from Passport chart
shots taken in the forest (greenish light), the results looked identical to when I
applied DCPs made from shots taken in clear daylight, or even indoors in overcast weather.
Either the difference is there but not visible in my small test sample, or the advice
give in the forum about it having to be clear daylight with the sun at over 45 degrees
in summer blabla is simply not true.

Reported by entertheyoni on 2014-05-05 12:30:46

[Beep6581]

Well .. it was me who wrote all these blabla about winter and summer daylight etc and
it was a summary of what I had read in the web from respectful posters like Iliah Borg.
Some months later Stefan took shots under "better" conditions and it proved by the
measures that the "winter" profile was better than the "spring" profile !!! :) .
The only I can think of this is that I mixed the priorities among icc vs dcp profiling
..
Looks like x-rite makes a very good job adapting illuminations :) making use of Adobe's
more or less accurate d65 & A matrices and then tagging the built matrix with the appropriate
illuminant .. 

But here we talk about the calculated WB temp/tint in RT where we have no tags for
the illuminant and who knows if any adaptation takes place ??.
As we easily see the different colormatrices are one reason for significant changes
in calculating temp/tint. I recall a Panasonic camera which had much more problem with
underwater WB when used with Huelight's matrix (rawimage.cc) instead of the default
d65 ..

DrSlony, I wonder what will the result be for temp/tint if you use the matrix from
your "forest" dcp vs the default d65 vs the current "automatched" dcp's .. :)

Reported by iliasgiarimis on 2014-05-05 13:10:02