|
Do note that they're talking about taking the full reference signal (between basically 0 and 10.000 nits) and display it on a normal HDR screen (which for instance can take a maximum of 1.000 nits).
That being said, in figure 20 they show multiple lines for target displays with different brightness capabilities, and they show a blue line for a screen with 100 nits max, which is a nice SDR target.
I'm no math expert, so not really looking at that, but just the text and the graphs basically describe what looks like a pretty normal 'knee' to map the values? Which I boil down in my mind to 'they are recommending to clip signals lower than a certain threshold, clip signals higher than a certain threshold, and use a s-curve for the left-over signal so it looks nice with rolled-off shadows and rolled-off highlights on the target display'. Am I wrong?
This will not really show any detail that might be present in the signal of 1000 nits or brighter, it just rolls it off and compresses it in the upper region. Algorithms like Hable are ment to recover some of those details, right?
|