Blue cast in underwater ORF raws


(Peter Krause) #1

Using RawTherapee 5.4 I noticed, that opening some .orf files in either the file management or the edotor turns them completely blue. It seems like during hte first few seconds they are displayed correctly before the blue cast appears. Other raw processors open those files correctly. Also, I think, those files show a green tick indicating they have been processsed but they haven’t.

I am attaching a sample file. Oly_Mauritius_2018_110.orf (12.3 MB)


(Peter Krause) #2

Sorry, forgot to mention: I am using RT with Kubuntu 18.10.


(Morgan Hardwood) #3

Both the white balance tool and the auto-matched tone curve tools fail on this underwater photo due to difficult lighting.

Using “Standard Film Curve - ISO Low”:

You can easily fix the color cast using the channel mixer, then adjust the white balance manually:

underwater.pp3 (642 Bytes)

I attached a partial processing profile underwater.pp3 which you can apply to your underwater shots.


(Thomas) #4

Manual spot white balance works well in darktable:


#5

Which version of Rawtherapee are you using ? In mine maximum chanel mixer value are caped at 200 and I saw in your pp3 that you have value up to 2100 ?

I’m using RawTherapee, version 5.4 (result of the rawtherapee -v command)


(Morgan Hardwood) #6

It’s the latest development version, which will be released as RawTherapee 5.5 next month.


(Morgan Hardwood) #7

Aim for these values:
screenshot_20181101_092412


(Peter Krause) #8

Hey guys, Thank you very much for your support. It helped a lot. I finally managed to to adjust the colors in RT.

As I said, other raw converters do not show this kind of ‘clipping’ effect. Hopefully, the next release provides an improvement on this issue.


(Ilias Giarimis) #9

We will have a correction in next release … :wink:
We just need to insert correct data in camconst.json … usefull would be if you could provide a set of raw samples as described in http://rawpedia.rawtherapee.com/Adding_Support_for_New_Raw_Formats sets 1 and 2 …

Dcraw has a wrong low “white level = 3345” defined for EPL2 instead of a correct around 4050 … so many reliable data are clipped and are not used for WB calculations.

Inserting in camconst.json (just before the Olympus E-PM2 paragraph …)

{ // Quality B, white level correction
    "make_model": "OLYMPUS E-PL2",
    "global_green_equilibration" : true,
    "dcraw_matrix": [ 15030,-5552,-1806,-3987,12387,1767,-592,1670,7023 ],
    "ranges": { "white": 4040 } // nominal 4056
},

takes care of this and then the WB calculations work (at their limit :slight_smile: )

For underwater shots it’s recommented to use the “underwater WB presets” first before applying autoWB … for this specific shot “underwater2” works best for me …


(Desmis) #10

If you choose in “White balance”
Underwater1 or underwater2 image is near quite good
After you can change temperature, tint or blue/red equalizer, or use after as says Ilias “auto”

jacques


(Morgan Hardwood) #11

@ilias_giarimis will you open a PR?


(Peter Krause) #12

Just uploaded a ZIP archive containing orf samples as described in rawpedia. The URL is https://filebin.net/xof3aan8hm0mgcpr


(Ilias Giarimis) #13

Thanks … although low iso frames are not completely overexposed due to strong vignette at the wide end of the lens … as rawpedia says …“it’s easier if you do not use a wide angle lens”


(Ilias Giarimis) #14

Sure … but tomorrow,

it’s a bit complicated situation … there are differences from body to body … samples from web I checked show different WL especially for ISO200 :frowning: (so for those samples the Dcraw value looks reasonable …)


(Peter Krause) #15

Sorry for not having read the instructions carefully enough. I could easily provide a set of at least the low ISO frames or a complete set
if it helps.


(Ilias Giarimis) #16

No need for new samples for the vignette problem … I can use a central crop …

… but the problem lies elsewhere … it’s the inconsistency of e-pl2’s white level …

So, if it’s easy, please provide a new set … I just want to crosscheck the consistency of the same body from time to time … and maybe find a solution in exif metadata …


(Peter Krause) #17

I uploaded a new zipped set of files (raw_samples_2.zip) to filebin.net :

https://filebin.net/iucpud9xz13lx133

Hope the files ok for your investigations this time :blush:.