X1D Color Artifacts in the OOC JPEG Files

BostonBoy · July 18, 2017, 05:26:57 PM

Sunday night I stopped at a roadside to take same spontaneous evening shots with my X1D. Looking at the files at home I've noticed a strange green color cast in the OOC JPEGs - in about a dozen of them. The files are clean when developed with Phocus from RAW. I really can't explain what is happening here. The best explanation I could come up with is a color artifact as part of the internal JPEG compression engine, even in high quality setting. Personally I don't think it's Moiré - it's a single color. I have posted the details in another forum so excuse me for posting the link here since I don't want to repeat myself:

https://www.dpreview.com/forums/post/59841220

I decided to re-post here since I got not response or explanation in ~24 h+ after my posting, so either people think it's not important since it's only present in the 12Mpixel JPEG "contact sheets" or nobody has a good idea on how to explain it either. This issue is legit, I have all the files and love my X1D, but I'm trying to understand what I did wrong to avoid the situation in the future. The other thought which came to mind is that I have never seen anything like it with Firmware V1.15, V1.15.1 or earlier, but that's pure speculation on my end, there are simply not enough data points. I have taken approx. 1500 exposures so far - never seen this problem.

Anybody else has seen this in their OOC JPEG files? Any good explanations you guys can come up with?

I'm only posting one example - look into the other thread for more details. Thanks for your Inputs.

NickT · July 19, 2017, 08:06:07 AM

Hi Ralf
I will send this direct to a Swede I know and see what he says, stay tuned!

BostonBoy · July 19, 2017, 08:23:16 AM

Thank you Nick!!

pahciao · July 20, 2017, 01:42:30 AM

Hi,
This is interesting. Could you post the RGB histogram of the RAW file? I am guessing channel clipping.

BostonBoy · July 20, 2017, 04:30:47 AM

That's a really interesting thought. I look into it when I get home. Thank you!

BostonBoy · July 20, 2017, 05:19:58 AM

Well, I don't think it is channel clipping. I looked at the RAW files in both LR6 as well as Phocus V3.2.1. Come to think of it that would not explain why the RGB to JPEG conversion looks fine. Attached are 3 files: once again one of the JPEGs (in 2000x1500 Pixel resolution), and the corresponding RAW file in both LR and P. Good thinking though!

pahciao · July 20, 2017, 01:34:26 PM

Hi,
Agree, it doesn't look like channel clipping from the RAW file histograms. In camera JPEG conversion often involves the application of a certain profile and curve prior to JPEG production because RAW files are generally too flat (that's almost the definition of a good RAW file!). Out of academic interest -- any chance to see the histogram of the RGB JPEG with the green streaks?

BostonBoy · July 20, 2017, 08:27:40 PM

Sure. Easy enough since I did not format the SD Card just yet. Attached. It does not look suspicious to me either, so I more and more come to think that it is either a specific gradient curve applied in the 16Bit (well 14Bit) to -> 8 Bit JPEG conversion, or a compression artifact from the tuning of the JPEG engine. What other explanations should their be? To some extend it would be easier if the RAW files and JPEG would show the same results, but it is clearly something happening in the conversion process. Would you agree?

pahciao · July 20, 2017, 10:17:12 PM

Certainly looks like something to do with the in camera RAW conversion and the tone curve applied. Does it only happen in the edge cases of images where one channel dominates?

BostonBoy · July 21, 2017, 02:02:02 AM

I can't tell yet, I simply don't know. I might have to run some structured tests to see what is happening, and on which of the channels. To some extend it's interesting that it seems only the Green Channel is involved, after all the standard Bayer pattern has 2 G, 1R, 1B Pixel, but I simply do not know enough about the algorithms involved demosaicing the interpolated colors etc. etc. - I'm just an engineer by trade, but in an entirely different line of business.

Maybe there is some tuning required in Sweden on the JPEG engine, I simply don't know. Maybe I did something wrong.

It will be interesting to see what the explanation is once the puzzle gets solved. I'm sure it will get solved, eventually.

NickT · July 21, 2017, 11:02:22 AM

It might be a while before I hear back from Sweden as they are all on holiday but I will post here as soon as I hear.

Beever11 · July 21, 2017, 02:03:31 PM

Got the same problem here couple of days ago. Waiting for a feedback
Thank you NickT

pahciao · July 22, 2017, 04:36:53 AM

Quote from: BostonBoy on July 21, 2017, 02:02:02 AM
I can't tell yet, I simply don't know. I might have to run some structured tests to see what is happening, and on which of the channels. To some extend it's interesting that it seems only the Green Channel is involved, after all the standard Bayer pattern has 2 G, 1R, 1B Pixel, but I simply do not know enough about the algorithms involved demosaicing the interpolated colors etc. etc. - I'm just an engineer by trade, but in an entirely different line of business.

Maybe there is some tuning required in Sweden on the JPEG engine, I simply don't know. Maybe I did something wrong.

It will be interesting to see what the explanation is once the puzzle gets solved. I'm sure it will get solved, eventually.

Hi,
Was watching the Karl Taylor short videos on the X1D, and he mentioned that the X1D captures RAW in Lab*RGB color space. JPEG would be either aRGB or sRGB depending on what you set. If your JPEG setting is in sRGB, the green bits could be out of gamut colour mapping by the in-camera JPEG engine, as sRGB would be a much smaller colour space.

BostonBoy · July 22, 2017, 10:54:22 PM

Hmmm...indeed the camera was set to sRGB and not to the wider Adobe RGB color space. Not sure if this was by accident (me) or a reset after the V1.17 firmware upgrade. I set it to aRGB going forward. Now - ideally I would like to re-process the JPEG from RAW now like on my Fuji S2Pro to see results...firmware backlog request I guess.

I take more pictures and report back. If this is the solution than Hasselblad still has to tune the gradient curves for the sRGB conversion though.

Thanks for your Inputs - looks like I missed the obvious.

BradP · July 25, 2017, 08:52:33 PM

Hope that works out. If it doesn't, I see all RGB channel highlights are slightly clipped in your histogram. That may be introducing color cast issues as I might be seeing in some RAW images I shot yesterday. The X1D firmware doesn't yet have highlight acquisition warnings (live view RAW (better) or JPG histograms or blinkies), and that may be contributing to artifacts. Just a thought.