Argyll for ICC profiling gives horrible results

[dje]

Ted I think I was a bit loose with my terminology. I don't think it has to be the individual RGGB channel raw data, it can be demosaiced into RGB data per pixel but it should be pretty clean of any other adjustments. If using the basic raw channel data as I did, you get the choice of using one green channel or the average of both. There's not much difference of course.

Dave

[dje]

Tarek

A couple more options.

This link gives a procedure to create a suitable tiff using Capture One. This has no color correction. It is actually a procedure put out by XRite for their camera calibration software but the first part about preparing a tiff should apply equally to using Argyll and a SpyderChecker.

The other thing you could look at is DCAMPROF which has a lot of information about Capture One icc profiles. It is based on Argyll.

Dave

[TarekAH]

thanks all for the replys...I am trying to go through them right now

I wanted to mention only that for a matter of fact, I could create an ICC profile that gives me really low DE00 (DeltaE 2000) near 1.05 and the colors look ok-ish (some washed-out effect could be noticed though all across the corrected image). HOWEVER, to me inspecting the ICC profile itself (gamut & tone curve) doesn't look "healthy" at all! and I assume no ICC profile must look like such or am I mistaken?

here is a preview attached of the "correct" ICC profile that presumably gives nice output despite its ugly shape
===========
so to sum up, my question would be:
What do you really think of such profile inspection (gamut/tone curve) is it anyhow acceptable at all as long as it does the job? or undoubtedly there is something wrong going on and must be corrected (this is what I think)?

[TarekAH]

What device are you trying to profile?
How did you create the tiff file of the target and most importantly under what condition?
Finally, where did you get the cie values for your specific target?

- PhaseOne Camera
- Libraw converter (I know that I can use PhaseOne Engine for the matter but for some personal reasons I need to use Libraw for this one + am afraid that it could be the origin of the problem!)
- the values are the one provided by Argyll itself....I have a real spectrally scanned values of my target and I tried to update the .cie .cht file accordingly and yeah a bit better but the core problem (gamut and tone curve shapes) still persists

[TarekAH]

Meanwhile, how about making an ICC from this:

http://kronometric.org/phot/color/Co...Bruce_sRGB.tif

It's a perfect MacBeth 24-patch TIFF created by the well-known Bruce Lindbloom and is dead accurate.

(no camera or converter involved) ...

Will it come out right, or be just as bad as the Spyder -that might tell you something ...

thanks @xpatUSA!
well...it works perfectly actually and this is what I call a healthy acceptable ICC profile (attached )
> so the problem is my original photo :?

[TarekAH]

This link contains my profile together with the .ti3 file created during the process i used. You could try this with your Argyll script. Also included is the reference data used for the Spyder patches (.cie file). Note that the Spyder patches are similar in color to XRite but not the same, and as you say, the order is different.

Some food for thought for you!

Dave

thanks @dje!
but looking at your profile's gamut it is still weird for me with that spaceship shape ... I assume no profile on earth should look like that ever, as the shape itself implies saying there is something wrong....don't you think?

[TarekAH]

Tarek

A couple more options.

This link gives a procedure to create a suitable tiff using Capture One. This has no color correction. It is actually a procedure put out by XRite for their camera calibration software but the first part about preparing a tiff should apply equally to using Argyll and a SpyderChecker.

Dave

well...even converting the IIQ raw image using COPE engine (CaptureOne) with no color correction and linear curve still the resultant profile by Argyll is so weird (the gamut seems to play some gymnastics ,check the attachment).... I am just getting really confused more and more why! :/

[xpatUSA]

well...even converting the IIQ raw image using COPE engine (CaptureOne) with no color correction and linear curve still the resultant profile by Argyll is so weird (the gamut seems to play some gymnastics ,check the attachment).... I am just getting really confused more and more why! :/

Tarek, I admire your persistence!

Here are the tone curves and gamuts together to make comparison easier:





I'm glad you tried Lindbloom's Macbeth TIFF. Logic tells me that there is something wrong with the TIFFs that your system is producing. I know nothing about the Phase One camera and even less about Capture One.

I speculate that IF the Phase One can produce a JPEG then shoot your Spyder card again, convert the JPEG output to TIFF and try Argyll again. I feel that the ICC might look more normal and would confirm that the problem lies with the previous TIFF files.

I'll continue to follow this thread out of interest but I have little to offer as to the root cause of the problem.

[xpatUSA]

I speculate that IF the Phase One can produce a JPEG then shoot your Spyder card again, convert the JPEG output to TIFF and try Argyll again. I feel that the ICC might look more normal and would confirm that the problem lies with the previous TIFF files.

I gave in and downloaded the TIFF you linked to in the OP. 30 minutes later I opened it up in RawTherapee and noted the six primary/secondary L*a*b* patch values and also the mid-gray values. Then I compared them to the listed Macbeth card values in a spreadsheet. Although they vary somewhat, the variation is not enough to explain the wacky Argyll ICC profiles you're getting, IMHO.



The difference between the Spyder and the Macbeth is that the colors are flipped with respect to the grays.

So I flipped them to be the same as the Macbeth and you might want to give this a try as if it were a Macbeth:

http://kronometric.org/phot/post/CiC...v-Macbeth.tiff

I predict that the ICC gamut will look tolerable. The tone curve might get a black level because the mid-gray patch is too dark.

I did notice that your TIFF is in Portrait orientation ?!

My opinion is that the TIFFs are fooling Argyll because of the patch locations.

[xpatUSA]

Ted I think I was a bit loose with my terminology.

No problem, Dave.

After all, this is a photography forum ...

[dje]

thanks @dje!
but looking at your profile's gamut it is still weird for me with that spaceship shape ... I assume no profile on earth should look like that ever, as the shape itself implies saying there is something wrong....don't you think?

Yes there's something not right. Out of interest, here's a Phase One profile for my Sony a7 which looks well behaved! I'll plug on a bit with this and let you know if i get anywhere.
Dave

[dje]

Tarek I played around further with this and produced another profile from my .ti3 file (derived from your raw file), this time using DCAMPROF. This free command line software uses Argyll to some extent but I'm not sure how much. I got much better results this time. The new profile is called "Dave.icc" and can be found in the same dropbox link as before.

The gamut diagram looks much better. I can supply my DCAMPROF code if you would like it. There is a special section in the DCAMPROF documentation on Capture One profiles and tone curves which is well worth a read.

[TarekAH]

thanks @Dave @Ted for the efforts, it is appreciated

Actually yesterday after many hours of getting helpless and as @XpatUSA has suspected that the problem lies within the TIFF image just then I realized that the actual problem is maybe starting to happen at the moment of raw conversion (Libraw) and well...it is indeed!!

I got to discover that the arguments of the raw converter engine are messing up with the resultant pixels apparently and specifically the argument "-W: don't auto brighten the image"

you may check the difference with and without this little freakish W argument (even though theoretically for accurate ICC profile creation this option should exist but it is making troubles!)

[dje]

Good to see some progress. I'm not sure how the auto brightening would affect it. Would it be possible for me to see the two .ti3 files for the above, and your improved profile?

Dave

[xpatUSA]

Good to see some progress.

Indeed -good to hear, Tarek!

I'm not sure how the auto brightening would affect it.

Perhaps a good question for the LibRaw guys Alex or Iliah on their blog?

Slightly OT: since I am not a Pro and I don't need absolute color accuracy -I gave up messing with ICC profiles long ago -especially when it came to creating many to suit all the possible scenes I might encounter!

So these days, I select the WB in-camera and work with that in post.

Occasionally, I'll get a Custom WB off a gray card if necessary (e.g. mixed LED/fluorescent at my bench or flash/LED in my kitchen) and that works quite well especially with my Sigmas which create a 3x3 matrix for Custom WB.

[xpatUSA]

thanks Dave and Ted for the efforts, it is appreciated

Actually yesterday after many hours of getting helpless and as @XpatUSA has suspected that the problem lies within the TIFF image just then I realized that the actual problem is maybe starting to happen at the moment of raw conversion (Libraw) and well...it is indeed!!

I use RawDigger which I believe uses LibRaw to produce an RGB review image which can not be adjusted i.e. RawDigger is not an editor ...

... BUT, it does have an auto-brightening option both for viewing and for export. That means that I could shoot my MacBeth card and export two RGB TIFFs -one autobrightened and one not. Then post them here for y'all to check. Let me know.

One thing I notice with RawDigger's export scaling which may or may nor relevant is that, if there is even just one super-exposed (e.g. stuck at max) pixel in the raw capture, the resulting export can be extremely dark. Whether LibRaw does that I don't know but ........ get my drift?

[TarekAH]

Good to see some progress. I'm not sure how the auto brightening would affect it. Would it be possible for me to see the two .ti3 files for the above, and your improved profile?

Dave

https://we.tl/t-AUR7ZRzcHf

[TarekAH]

I use RawDigger which I believe uses LibRaw to produce an RGB review image which ...

thanks Ted.
if RawDigger underlying system is the same as Libraw, basically using it, then I presume that we should get the same wacky behavior whenever the auto-brightening is active.

[xpatUSA]

thanks Ted.
if RawDigger underlying system is the same as Libraw, basically using it, then I presume that we should get the same wacky behavior whenever the auto-brightening is active.

I would think so too. <>

Unfortunately, RawDigger has auto-brightening grayed-out for RGB exports so I must withdraw that suggestion, sorry.

I could do it with DCRaw but that is too far removed from LibRaw even though LibRaw is based on DCRaw ...

[dje]

Hi Ted and Tarek

I came across this article by Norman Koren of Imatest (and Gamutvision) fame. It talks about how gamut diagrams for camera profiles are of limited use as they don't represent the gamut of the camera system as a whole. Have a look at the first paragraph and the last section entitled "Response to LULA's Leica M8 Review" in particular.

This doesn't mean that we shouldn't be concerned with very strange shapes looped back on themselves but that we must remember that we aren't getting the full picture from such diagrams alone.

I have a feeling that the wider gamuts probably correspond to proflies that are not tightly optimised (very low DeltsE's) for a group of 24 color patches but I would like to do some further investigation of that.

Dave