I don’t think it’s possible to simultaneously compress the dynamic range, depict the magnitude of the brightness difference, and also not have a halo… it’s just a matter of deciding which of those you want to prioritize.
You can control how wide the halo is, and maybe I should have made it a bit wider, but that depends on the viewing distance for the image.
Agreed for an image like this you have to make decisions…I think just running back quickly the edits using ART and RT seemed less prone to the halo on the trees. Personally I like this image with the dark foreground or in silhouette I think that makes the most out of the sky…but perhaps some aspect of the fog or the color in the field in the foreground are the focus…every ones eye lands a bit differently …
I’d say halos appear whenever you have two regions having different exposure correction. Somehow you have to transition from one to another. Either there is a sharp transition (which usually leads to dot-like artefacts, when some pixel on the boundary gets exposure corrected using the value from one of the regions, while its neighbour, with an only slightly different value, using that of the other region), or a gradual one (which always leads to halos of some sort – smaller, but harsher, or larger, but more gradual). As Todd put it:
Handling the input camera profile differently could be a part of it?
Adobe probably interpolates between the two calibrated color matrices based on the selected white point (they did write the DNG spec after all, and presumably use the equivalent two-illuminant DCP profiles; furthermore, they can also use LUTs there as well). dt doesn’t and always starts from the D65 color calibrated matrix and does the white point adaptation later.
The interested can see the “Mapping Camera Color Space to CIE XYZ Space” section of the DNG spec for more details.
Thanks for your posting! And thanks everyone for such smart and interesting contributions!!
I’ve just tried to see which could I get in 3 to 5 minutes as much, and using tricks or methods I do use frequently, so that this could be more or less my usual editing for such a picture. In dt 3.8, of course.
I find highlight reconstruction module being more a problem than a help this time, so I disabled it and used Filmic’s own reconstruction system. A fast preset in Tone equalizer (not even adjusted it). And some contrast in multiply and Colour Balance. That’s it mostly.
It’s not the best edit here, of course, so many great renditions! But just a try on what can be get in dt nearly as easy as LR. And I have to say that I don’t feel too bad about result.
I don’t know I that adds anything to the discussion, but that are my 2 cents…
Unlikely. Color profile works on individual pixel intensity. Highlights reconstruction works on pixel neighbourhood and trying to infer the original values of pixels.
Oh, I didn’t mean that for the highlights, but as the comment on the general “dull” impression some people had…
I don’t know if it would be applicable to this particular image, but a trick I tried the other day to eliminate (reduce?) halos was to apply the offending module with a negative details mask. That way I could partly decouple the aggressive values of the module from the steepness of the transitions at the edges in the image, playing with the blur mask parameters.
When you hit the bright sky - dark tree boundary, if you use a negative details mask, it won’t affect that edge so hard. So, if you’re darkening the sky, then just around that edge it will be darkened less; if you are brightening the trees, they will be brightened less.
It’s worth a try, but I think the compromise remains.
Of course, there is still a compromise. But now you have a couple of degrees of freedom more to play with.
PK1_3540.DNG.xmp (15.4 KB)
When I first started working with DT I felt the colors were more dull than what I was used to from the out of the camera JPGs or LR edits. But, RAW files are dull. Just set the camera’s picture style to neutral to see this. It is the choices made in processing that replace this dullness with the color and vibrancy that we tend to like. DT doesn’t enforce this upon. LR knows most people like vibrant pictures, so its basic processing produces such a look. The SOOC JPG images are an interpretation by the camera manufacturer to produce images that will be liked by the masses. People confuse this look with the correct rendition of the scene.
I have come to learn from experience how to produce images that are not dull and match my desired interpretation of the scene. Filmic offers beautiful control over the contrast of the image without clipping the highlights or the shadows, however, the color may be left a little dull. I have learnt to tackle that using the ‘color balance rgb’ module. There is a preset call add basic colorfulness and this puts the punch back into the colours. I have made my own preset for add basic colorfulness 50%, where I have positions the sliders at just half the values. This module is very standard part of my workflow to produce the picture style that pleases me. There is no reason for accepting dull images out of DT. All that being said, I still prefer how LR handled this specific image from Christian. I just wonder what is in Adobe’s black box that pulls out the clouds that are hidden in the extreme highlights.
RT dev. I couldn’t figure out why I had some artifacts in the sky, turned out it was the automatic lens vignette correction that was messing with highlight reconstruction somehow. Not sure if that’s a general problem but good to remember.
PK1_3540.DNG.pp3 (30.6 KB)
Good exercise here.
In my rendition, the key modules were highlight reconstruction and exposure, and getting some detail back with a masked contrast equalizer and also color balance rgb.
In highlight reconstruction, I first thought reconstruct color mode was going to be it, but I got the vertical and horizontal lines. I went with reconstruct Lch but later after messing with all the other modules I found that even clip highlights provides a good result, and I was unable to reproduce the vertical and horizonal lines.
Mine has a distinct pink patch in the sky while others have a nice gradient.
Lastly, when exported, my jpeg (my first one below)looked like garbage. never had such a difference bettween the export and the look inside DT. I turned off the 2nd contrast eq and exported again for the second jpeg.
PK1_3540.DNG.xmp (9.4 KB)
this one is really good, I studied your XMP. I like how you used the new diffuse or sharpen module. I have no idea what do to with it. The only think I’d change would be to bring up the middle tones on the tone eq to add more definition to the trees in the fog. They are really washed out right now.
Hi, thanks for your reply. I mainly focused on the over exposed part in the sky trying to get it fixed.
I tried various highlight recovery reconstruct settings. The reconstruct in color looked best to me, but I a gird showed up. So I took inspiration from a couple of other posts here.
Disabling the highlight recovery module and trying to use the highlight recovery part in filmic. By gently moving the slider to grey I hoped blending in the parts. There still is a magenta ‘blob’ visible, but it’s starting to get acceptable imho.
The diffuse and sharpen module is also very new (to me) I just used the built-in presets.
I’ve watched the videos by Aurelien Pierre, but I have to put my time in it to fully understand how to use them. Since this module exists I haven’t used the local contrast module or contrast equaliser…
Nouvelle version entièrement traitée avec Art
New version fully processed with Art
PK1_3540.DNG.xmp (15.9 KB)
