Removal of magenta highlight artefacts

Didn’t read past the first few sentences (baby on my lap ) , but something did click in my mind :

The reconstruct options clip and lch try to produce white . And if color calibration comes after that to apply something, it tints the white.

I was comparing legacy vs modern with the reconstruct module off, and then you see no difference.

And since i only use reconstruct color or segmentation in practice, i don’t have the tinting issue . Those two don’t try to produce white, they try to interpolate from what’s around, no matter the current white balance situation.

So i kinda stand by what i said, modern or legacy makes no difference in how clipped highlights look. It does however differ in how reconstruct clip or reconstruct lch look. And since clip is still the default AFAIK, people blame it on this ?

That’s not completely correct. While estimating what’s around they use the other color channels. This estimation is much better, if good coeffs related to white are available.

1 Like

I understand. But in practise i haven’t seen a case where it mattered .

Well, I must say this issue is one of the highlights reconstructing problems that’s driving me crazy!

3 Likes

And also what it looks like when filmic compresses those highlights which is where the tinkering can start. The good thing is all the options are there for the most part but you need to craft your own if you want to be free of it from the start… In you case the mode you use for filmic and segmentation must be working for your eyes and giving you the result you expect… which is all that matters. My take home lesson from recent weeks is confirm your raw white point and do some work to test and evaluate your D65 defaults or actually calibrate if you have the means and then this side effect of less than accurate D65 values will be less of an issue…