highlight reconstruction issue

test_5290185.jpg.out.arp (10.5 KB)

Done in ART 1.8.1 (a rawtherapee fork), just 1/2 dozen clicks almost straight out of the box processing. The main modules of the edit are the Tone Equalizer and the Dynamic Range Compression

Syv

Interesting how this has become very much a #processing:playraw. @eyedear It may be time for you to clarify the photo’s license; e.g. one provided by the Creative Commons.

Filmulator handles the shoulder area really nice.
There are still some magenta amobae on the clouds, i would simply get rid of them in gimp.

3 Likes

Darktable 3.4

Girl_blown_highlights.orf.xmp (10.1 KB)

This is the embedded jpeg from the raw originally posted. I think it contains the mosaic that the original poster mentioned. If this is the case, does that mean it has nothing to do with darktable?

The jpeg was extracted using the excellent Fast Raw Viewer.

Those seem like compression artifacts.

Offtopic, but with such a gorgeous model, I couldn’t resist to make another retouch, this time concentrating on the subject :slight_smile:
Also used Portrait Pro and Gimp

2 Likes

you fixed the blown spot at the right shoulder well

1 Like

Just used the clone tool in Gimp

2 Likes

I think there’s problem with color calibration module that is above highlight reconstruction. It tints reconstruction result, so I had to reverted into legacy mode with WB only.
I lack color correction skills but except demoisac artifacts I think that result can be quite good in DT with white clouds.


_5290185.orf.xmp (11.1 KB)


_5290185.orf.xmp (29.3 KB)

3 Likes

@age think that @Wayne_Sutton has posted those artifacts here https://pixls-discuss.s3.amazonaws.com/original/2X/6/6e58ef2f78bb2f2be9e61ff170342ff9cbb3d09a.jpg
Indeed LCh is wontfix but this is color mode, or am I wrong?
Maybe this bug (Bug #10283: new “reconstruct color” in “highlight recovery” producing weird artefacts outside highlights) hasn’t been fixed and bug report was lost after migration into github…

2 Likes

@eyedear can I use this image to report “moire” issue?

Great image for highlight reconstruction algorithm development!!

Thanks so much for sharing.

Result of my algorithm and experimental raw converter (saturation 1.5x WB 4700K)

We are fully aware of severe problems with dt’s highlights reconstruction algorithms.

I’m working on this, doing something new in collaboration with @iain from gmic. It’s a pretty difficult thing to do but preliminary results are very promising. Nothing for next 3.8 version unfortunately…

12 Likes

You make me curious :slight_smile: Good news !

I think @iain posted something going back a bit about inpainting highlights???

EDIT

I think this is what I was thinking of…

Exactly. Still some major points to solve but there is steady progress and performance seems to be manageable. Cross fingers …

8 Likes

That’s great news. Hope that this would make it to DT in the future

ART