Dual ISO causes funny colours

[pletharoe]

After post processing with CR2HDR the colours in half of my dual ISO shots are off.  Increasing the temperature and tint in DxO seems to fix it.

Any idea why?

[Walter Schulz]

cr2hdr version? Camera type? ML version? Downloadable CR2?

[pletharoe]

CR2HDR version: Unknown... downloaded a couple of days ago from https://builds.magiclantern.fm/jenkins/view/Other%20tasks/job/cr2hdr-20bit/

Camera: 5D3

Firmware: 1.2.3

ML nightly build: 2015 April 28

Files can be accessed at:
https://drive.google.com/folderview?id=0B_8z7XiBut6rcG9hZVg4SFZRWGs&usp=sharing

[a1ex]

That's a known limitation - just set WB manually in your image editor.

With cr2hdr, by default, white balance is not the one you set from camera; instead, it attempts to compute WB on a large surface that appears to be gray or white. In your first image, it found the sky as the reference gray (wrong choice), and in the second image, it found the baby tray (probably correct).

You could try the option --wb=exif to copy Canon's WB settings, but last time I checked, it wasn't working very well.

More details here: http://www.magiclantern.fm/forum/index.php?topic=10265

[Danne]

I,m a little curious if it would be good to use the wb multipliers coming from the dual iso CR2 file?

Wth dcraw you get the numbers

Code: [Select]

dcraw -i -v  DUAL3755.CR2

Camera multipliers: 2032.000000 1024.000000 2009.000000 1024.000000
Dividing gives following. (Thanks Chmee)

1024/2032=0.50393701
1024/2009=0.50970632

Code: [Select]

exiftool "-AsShotNeutral=0.50393701 1 0.50970632" DUAL3755.DNG


Original cr2hdr20bit


dcraw multiplier wb from CR2 file


With command line set --wb=exif (seems to give the standard 5500/19 temp/tint)


Original file
https://drive.google.com/file/d/0B4tCJMlOYfirVmRQVXhULWNvcDQ/view?usp=sharing

[a1ex]

Sounds interesting.

I tried on an image file taken with the following WB settings, on 5D3:

Code: [Select]

White Balance                   : Manual Temperature (Kelvin)
Color Temp Kelvin               : 3100
WB Shift AB                     : 0
WB Shift GM                     : 0

With dcraw/exiftool:

Code: [Select]

Camera multipliers: 1431.000000 1024.000000 2497.000000 1024.000000
exiftool "-AsShotNeutral=0.71558350803633819 1 0.41009211053263916" TEST.DNG

Result in ufraw: 3135K, green=1.051.

Pretty close, and... the result is identical to cr2hdr --wb=exif.

The AsShotNeutral multiplies for 3100K/green=1 are 0.70488444 1 0.40021194. That would give multipliers 1453 1024 2559 1024.

[Danne]

Yes, around those figures I get too so it is very close. If it reads the auto white balance rather than grabbing kelvin settings it should be usable. See if I can find more pics. Left my cam at home.

[Danne]

Exactly the same. I, ll be damned. 

So I,d say the wb is close enough to be considered correct from camera.

Retested wb multipliers from two files from dfort and numbers matched exactly as wb=exif.

[pletharoe]

Thanks very much for the info and THANK YOU SO MUCH A1ex!  ML is awesome!

[Danne]

Tried another file.
Original here
https://drive.google.com/file/d/0B4tCJMlOYfirMXlnWUtuT2hVWFk/view?usp=sharing

Got all greenish looking wb from using --wb=exif(cr2hdr) whereas dcraw multiplier gives more or less correct whitebalance. Not sure what,s going on here?

dcraw multipliers

Code: [Select]

exiftool "-AsShotNeutral=0.50294695 1 0.51457286" /Users/itguser/Desktop/ml_vfs_2/DUAL2695.DNG

dcraw_multiplier(CR2)


wb=exif(cr2hdr)