Yes, this is a tricky image for highlight reconstruction. ‘Inpaint opposed’ fails because there are objects with different colours that are clipped and ‘inpaint opposed’ restores highlights with one colour.
I did have some success with ‘segmentation-based’ reconstruction.
In this case, the area near the out of focus hair is the correct colour. This is because the hair is blocking some of the light and reducing the exposure of the object behind it.
So, oddly, I got my result by raising the clipping threshold.
Changing the clipping threshold in ‘inpaint opposed’ or ‘segmentation-based’ reconstruction will affect the colour of the result. This is because both methods look at the unclipped pixels surrounding the clipped area to estimate the correct colour of clipped areas. Adjusting the clip threshold changes which pixels are considered clipped and therefore which pixels are used to estimate the colour of clipped areas.
@hannoschwalm you may be interested in this image for highlight reconstruction tests.
This shot’s lighting conditions captured my interest while moving to DT and now that I’m subbed to the forum here’s my edit!
Pretty basic stuff and adjustments (I suppose), the only thing worth mentioning is I wanted foliage to look as natural as possible while keeping both interior and exterior warm.
I focused on avoiding the electric blue-greens of the foliage, mixed with colour calibration and some masks + opacity to lessen the punchy blue look from outdoors.
I did try different approaches to colour calibration, but nothing I liked more than the Incandescent + drawn & parametric masks. I probably went overkill with them, but at least I learnt something new, and I will definitely try different approaches next time.
