Originally Posted by madshi
So here's a big new test build
This is amazing, I'm having a hard time seeing any difference between the "live" and stored measurements algo
The default settings are pretty good, but unfortunately there is still a difference for content with and without black bars to set the threshold (similar to the scene detection for the tone mapping target).
Two scenes I used for the threshold (most significant to me):
Pan, frame 93956 (01:05:19), real cut
- Lum changes (black bars included): 19.44%
- Lum changes (black bars excluded): 26.12%
The Meg, frame 1249 (00:00:52), false cut
- Lum changes (black bars included): 14.83%
- Lum changes (black bars excluded): 19.60%
Therefore, I got two answers for the threshold:
- With black bars included: 17.5%
- With black bars excluded: 22.5%
As for the max brightening/darkening per second, 50
is working fine for me on Infinity War panning shot at 26min.
I tried with 100, but there was some bad flickering on Passengers scene from frame 61249 (00:42:35).
I also looked at the sandstorm scene in Fury Road, but I don't see any problem there if I don't do a frame-to-frame analysis.
Edit: In addition to not having to measure titles anymore or having to try to manually find a static target (which didn't work very well anyway), I don't notice any performance drop