rawtherapee.com

svenl · Joined: 06 Jan 2008 Posts: 6

I think I found a bug in the handling of the color management input profile when working with RAW images. I have a Nikon D50 camera, and up to now I got slightly too dark images with several RAW Converters, including RawTherapee, until I realized that I did not use the input icc profile for my camera. With the profile configured into the RAW Converters, I get correctly exposed images e.g. in UFRaw, but in RawTherapee the images get even darker. Could it be that the input profile is applied the wrong way round in RawTherapee? I am using version 2.2.

Regards,
Sven

SturmMD · Joined: 28 Oct 2007 Posts: 18 Location: Minneapolis, MN USA

I am experiencing the same type of problem with my Pentax K10D and a custom profile I made using lprof.

I am hopeful that this can be resolved because I really liked the results of when I manually applied the profile using photoshop

keenonkites · Joined: 24 Aug 2007 Posts: 417 Location: Switzerland

SturmMD, svenl,

Gabor is very soon releasing rawtherapee 2.3 with a new features and bugfixes, aswell as using new dcraw engine...

can you please test if it behaves the same in the 2.3 Release Code also ?
here the post containing the RC:
http://www.rawtherapee.com/forum/viewtopic.php?p=2292#2292

Thanks and regards
Patrik

DrSlony

Release Candidate

absolution · Joined: 17 Nov 2007 Posts: 321 Location: Prague, Czechia

The profile you are applying in RawTherapee was created using RawTherapee or UFRaw? Lprof documentations explains how to create profile using UFRaw but I suppose this profile could only be used in UFRaw then.

As for my experience the camera icc profiles could'nt be shared between applications (yes, this should'nt be but it's just like that). I have ICC profiles for Pentax *istD from Bibble and Rawshooter and these are totally different and both unusable in UFRaw and RawTherapee.

I think you would need to create profile using RawTherapee but I'm not sure how or even if it's possible. This is because the RawTherapee lacks (or I think it lacks) unmanaged color input setting (something like color matrix turned off in UFRaw).

But maybe someone more familiar with RawTherapee profile handling could explain.

svenl · Joined: 06 Jan 2008 Posts: 6

The problem still exists for me with RC2 Sad

The profile I am using is the original camera profile available from Nikon. I think any profile should work the same way in every software, shouldn't it?

Regards,
Sven

keenonkites · Joined: 24 Aug 2007 Posts: 417 Location: Switzerland

DrSlony,
... uups... thanks.. Wink

svenl,

thanks for the tests, allthough I've to say a different result would make me happier.... Wink

I would also guess that the same profile should behave the same in different applications as long as it's not combined with other features of the specific software....

Unfortunately I can't help you any further here, we've to wait for a possible feedback of Gabor starting next week (most probably)...

Patrik

absolution · Joined: 17 Nov 2007 Posts: 321 Location: Prague, Czechia

I think when digital cameras are concerned color profile is always more or less software specific. The reason is that there are couple of program specific steps before the profile is applied - decoding, demosaicing, white balancing...

I would expect that even different demosaicing algorythms would result in slightly different color rendering.

svenl · Joined: 06 Jan 2008 Posts: 6

Hmm, that might be true. But I still think that RawTherapee should not behave opposite to other RAW converters when applying an input profile, making images much darker where other converters produce brighter images with the same input profile specified. Unfortunately I do not have an IT8.7 chart, otherwise I would try to create my own input profile with RawTherapee and LProf, to see if RawTherapee would apply such a profile correctly.

absolution · Joined: 17 Nov 2007 Posts: 321 Location: Prague, Czechia

svenl · Joined: 06 Jan 2008 Posts: 6

Ok, it seems you're right. I just made a quick test with a profile I created with LProf to drastically reduce red color values (I manually set red gamma to 4.0, green and blue gamma to 2.2). Now I get greenish images in both UFRaw and RawTherapee, so there is probably everything ok. So I guess I have to create a special profile for the combination of my camera AND RawTherapee.

Regards
Sven

SturmMD · Joined: 28 Oct 2007 Posts: 18 Location: Minneapolis, MN USA

I created my K10d profile using lprof and a IT8.7 chart. I experience the "darkening effect".

My understanding of what was going on/is necessary is as follows.

1. I used RawTherapee to develop my raw image of the IT8.7 chart.
2. All the settings were to neutral.
3. I used the "camera" input profile and selected a large output color space (bestrgb).
4. Saved output as TIFF 16 bit
5. Using lprof and the color chart's reference file, I created a profile.

I then named this file properly (ex. "K10d from bestRGB")

I verified that the profile was good by opening up my 16-bit TIFF BestRGB file in photoshop. I then "assigned" my profile to it and everything was great: gray became neutral, colors were corrected.

But as noted before, when I chose this profile in RT the results aren't correct.

Now, any application of gamma during the initial development could be the culprit. This means that developed chart I analyzed in lprof had gamma applied but when I select my custom input profile the gamma is bypassed by RT. This is just a theory

Is there a way in RT or any other tool to get a direct dump from raw to tiff (no linearization, gamma, etc) with ONLY demosaic. This would probably be what is necessary to use in lprof.

absolution · Joined: 17 Nov 2007 Posts: 321 Location: Prague, Czechia

I think you should not use BestRGB as output color profile but No ICM: sRGB. I think this is what Lprof expect to get (Lprof documentation also say to use sRGB output for making profile in UFRaw ). Though I agree with you that this would be probably of no help, because as you note, we cannot get really color unmanaged result from RawTherapee. And it seems that it's hardly needed for creating a valid profile. Or something other is wrong...

svenl · Joined: 06 Jan 2008 Posts: 6

Just an idea... Could it be that RawTherapee applies the input profile at the wrong place in the image development process - e.g. after applying gamma where it should be before?

gabor · Site Admin Joined: 19 Mar 2007 Posts: 315

I think input profile handling is working as it should. If I process an image with dcraw and specify the input profile in the command line, the output colors of dcraw and RT are the same.

The difference between ufraw and RT could be that RT does not apply the camera matrix on the image.

SturmMD: there is no way to obtain the unprocessed raw colors with RT. Why would it be useful?