Maybe another question. What are ālinear srawā files ?
A compression format used by canon ? Why ālinearā ? Is it already demosaiced in camera to store the file as a different resolution or something?
It means the rgb data are already linear - or you could say already demosaiced. But they are not yet corrected for white balance.
I think fixes for these files will hopefully fix linear DNG HLR as wellā¦
I think and hope so. Since @hannoschwalm only wants out of camera and not linear DNG generated by other tools Iāll wait for his PR and test some of them.
If you have proper files generated otherwise, go ahead. (there have been so many crappy dng files presented in issues I feared to get āspammedā
DSC_4287_dxo6.dng (60.7 MB)
From Blowing steam and highlights :-)
20120116_D7K0879 (1)_dxo6.dng (57.7 MB)
From old dubai skyline (and highlights reconstruct in filmic)
Feel free to label them as unusable for you. That means my problems with linear DNG and HLR are unrelated to what you are working on .
These are DNGs that otherwise load and work fine. I can also export them in a way that the writing program does its own HLR, and then Iām perfectly happy with them. But Iām guessing the data doesnāt truly āstay linearā then (have to lower exposure in DxO before export) and while DxOs HLR is working, itās not the best in any way.
Iāve done as little as possible to the files. Should be pretty much as close to the original raw as possible, but demosaiced and rescaled to be between 0 and 65535 (so DxO interprets the black and white levels of the source raw file).
Pick one, or all, or noneā¦ all up to you!
(This file keeps saying there is an upload error. To big? (Over 100mb).
From To the Skies - Strange behaviour in highlights
Do you want me to open a ticket?
No. Already have it fixed here, just have to do the opposed means yet.
Thanks @jorismak - i knew those images as raf files so very good to see how it works out with these dngs! Those two work in a perfect way, could you share some more?
7D CC0
20221012_M7D6679.CR2 (17.2 MB)
20221012_M7D6680.CR2 (11.1 MB)
20221012_M7D6676.CR2 (15.5 MB)
20221012_M7D6677.CR2 (10.1 MB)
20221011_M7D6659.CR2 (14.5 MB)
20221011_M7D6660.CR2 (9.5 MB)
Anything you are looking for ? Large clipped areas ? All channels clipped , only a few ?
A selection of very strongly clipped, blown out areas surrounded by dark parts. Artificial lightingā¦
This is what comes to mind. The clipped areas are probably not surrounded by dark parts.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License .
(Hope I did the CC thing right).
_DSC0962.ARW (47.0 MB)
_DSC0948.ARW (47.0 MB)
_DSC0962_dxo6.dng (80.7 MB)
_DSC0948_dxo6_2.dng (81.0 MB)
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License .
(Hope I did the CC thing right).
This one is a large patch of clipped against hard edges of darker / āreal contentā. But not all channels are clipped, though.
_DSC0984_dxo6.dng (71.0 MB)
_DSC0984.ARW (46.9 MB)
Not at all what you asked for, but I always think this is a nice test case because there is a lot of different channel clipping happening, and every algorithm seems to produce something different.
P4290427.ORF (14.1 MB)
P4290427_dxo6.dng (52.4 MB)
Thanks, I will check.
Iāve noticed itās in master, so I made a new build to have a go.
First thing I noticed - which might very well be my own fault so none of your concern - is that I opened a DNG file, and Iāve seen a pop-up saying āhighlight method not availble, falling back to inpaint oppsedā or something like that.
Probably because I have a preset auto-enabled for setting segmentation with combine to 2 and candidating to 0% as a starting point.
So, nice that it tells me thatās not available and instead switches to oppsed. The thing is, like the right 2/3rds of the image was black. Once I disabled HLR and re-enabled it, it worked fine!
The normal Darktable clipping indicator seems to still do nothing.
But the threshold viewer in HLR seems to work fine.
If I then manually select my āsegmentation based defaultā preset again, it sort of activates it and also sort of falls back to opposed again .
I do see the popup for āfalling back to opposedā. My guess is if you fall back to opposed, there needs to be a āon changeā event triggered on the dropdown or something, for the interface to be notified of the change .
Well, the menu was wrong for linear dng or sraws. There never was anything else functional than clip now we have opposed in addition.
Arrgh, I see what you mean, will take care of itā¦
As you probably have noticed the segmentation algorithm is not yet implemented for sraw files.
I donāt know yet if the results are as good, would need more examples with smooth transitions from unclipped to clipped, all color channels should be involved.