Don't see what 'reconstruct' tab in filmic-rgb is doing

I have an image with a background which is a patchwork of many tiny areas of sky seen through a collection of trees. With the ‘raw over-exposed indicator’ toggled on, this patchwork appears pretty much uniformly red. In the ‘reconstruct’ tab in filmic I am using the technique of setting the threshold so that only these red area are visible in an otherwise completely black mask. So I assume that this has accurately defined the areas of the image which need reconstruction of the highlights, But what I don’t see is where I actually change something to effect that reconstruction - the patchwork stays as red even after setting the threshold, while I expected it to no longer show as ‘over exposed’. What is it that I am not understanding here?

Whatever you do, an area that is over-exposed in the raw data will stay over-exposed in the raw data. Afaik, that measurement is taken before the reconstruction happens.

Also, note that the highlight reconstruction module has an influence on the result of filmic reconstruction and that the filmic reconstruction is quite sensitive to the value of the threshold parameter

The raw overexposed indicator will always show the areas that were overexposed in the original raw image, it shouldn’t be affected by any processing later in the pipeline. If you turn off the overexposed indicator and move the threshold up and down it should be possible to see what reconstruction is doing.

I always have trouble with patches of bright sky through trees, so many things can go wrong there… purple fringing and chromatic aberration that can be exacerbated by filmic’s ‘preserve chrominance’, interpolation artifacts that occur at edges between bright and dark areas when the image is rescaled, and discontinuities between the middle of the bright patch that is overexposed and the edge that isn’t.

From Filmic’s author:


Aha - so it isn’t just me, then? I am somewhat surprised by the amount of purple fringing I see as, generally, both of cameras and usual lenses have been quite good in minimising the creation of these effects.

I have noticed that the Olympus’ built in jpeg processing is very good at suppressing these small artifacts around bright sky and trees, even if it can be a little heavy handed with the sharpening sometimes. I’m not really sure how to replicate whatever it is doing in Darktable.

@paolod Would it not be a good idea, then,
to offer an image for Play Raw, and indicate
where your problems might be?

Have fun!
Claes in Lund, Sweden


I have a similar or related question.

This is one part of the Filmic module I studied and really tried to learn. When I apply what I learned, I do see differences in the results, but the differences rarely seem to be improvements.

I know that I should start a new Play Raw with an example that just focuses on that aspect of processing, and what I see in the results. I will do this when I get the time.

PS: Here is the new thread Re: My question about Filmic highlight reconstruction