There are two blurs: one is for the mask, indeed, but the other, the bloom, is for the image.
The problem I encounter most often is color artifacts (hue shift) in highlight areas, which are only blown out in 1 channel. Typically this affects the sun and / or its surrounding. With APSC sensors in a single exposure this is often the case in challenging lighting conditions. And blowing out the sun is pretty common I think.
Maybe a few terms overlap here. Or, there are multiple issues, and you might confuse some⌠or you are having them all at the same time :).
A âhue shiftâ problem you read about, is that bright reds or yellows (note, bright colours, not really highly saturated, although they often go together). What happens is that the reds can turn a bit âsalmonâ and the sky colours can shift to something people not always want (in the yellows).
Iâd call these effects âsubtleâ. And then I donât mean that they arenât a real problem, but itâs not like âblobs of weirdnessâ or something that jumps out. Itâs a colour âthat is a bit off to what you expectâ.
This is caused by filmic v6/v7 and the way it tries to preserve the colours in bright highlights. Simple workarounds I read about are to use sigmoid instead of filmic, or use filmic v5 (with or without the preserve-chrominance mode set to ânoâ, Iâm not sure about this).
This has nothing to do (per se) with clipping sensor data! This just happens in bright colours, notable yellows and reds. And some people are more sensitive to it than others .
Sensor clipping produces different issues. Most notably, you will see that if you disable âhighlight reconstructionâ and âfilmicâ and all other modules, except âexposureâ and you keep lowering exposure: Youâll see how your highlight data really is captured in your file. Most often this is âthe magenta highlightsâ you will hear about, but it can also be cyan or some other colour (depends on the white balance).
This is not a simple colour that shifts a bit to another. This is really âthe wrongâ colour appearing in your highlights. This is what the âhighlight reconstructionâ module is for, and the ânewâ inpaint-opposed method (since 4.2?) is a real winner here. Thereâs a reason Rawtherapee included it quickly as well, even if their âcolour propagationâ algorithm was also very good :).
This module will try to fix those blown highlights by using data from the channels that are not clipped around it.
This has nothing to do with filmic (they appear without filmic enabled as well). But some algorithms (like sigmoid) turn highlights more to white or other forms of removing the saturation from the highlights, which just means those wrong colours are less visible.
The best way here is just to not blow your highlights. And this has nothing really to do with APS-C or sensor size (as someone shooting more and more m4/3 instead of full-frame, I can know).
It just means dialling in some negative EV when you are in bright surroundings. I often shoot with -1EV or -1.5EV on my m4/3 when itâs (harsh) sunlight outside.
Full-frame sensors clip just as easily, do not think thatâs a fix for it :). They can often handle the under-exposure better, but you still have to make sure you do not over exposure yourself.
Unless you have an (older) APS-C Canon DSLR, most APS-C cameras out there (even ones from 10+ years ago) can easily handle a few stops underexposure to preserve highlights.
Great content! Thank you so much for the elaborate explanation. This really gets some issues straight for me. I tried it on a real tough example and now at least I get rid of most of the hue problems and the sun doesnât look like a lightbulb either. Great result! I guess, I can refine this a little more, but this seems to be the way, I have been looking for.
@jorismak has provided a very good explanation/solution that I agree with. I would just add that what I like about filmic 7 is the new highlights saturation slider that can have a huge impact on sunset images and really all images with highlights. Move the slider right and left and just see the effect. @123sg I have used your recent sunset image here to demonstrate the new slider in filmic V7.
Iâm still stuck on v5 as my preferred option⌠I love all the control and the lack of the strict gamut mapping⌠I like to use the latitude with the saturation slider⌠that is when I use filmic⌠I bypass it much of the time unless the image is one with big DNRâŚ
Hi Todd, that is what I love about DT and filmic, we have the choice to use filmic v5, v6 or v7. You have previously shown me that v5 can be very different to v6. V7 is more intuitive to newbies and I find the slider sometimes work best in the center, sometimes left and sometimes right. I will still remember that 5 or 6 may be needed for some images and sometimes I prefer Sigmoid.