Originally Posted by Javs
Dude... I consider myself a fairly switched on guy, but seriously, reading your posts, I dont know if its you or me, but this stuff from you, every time its just SO difficult to read through... Is anybody else feeling this? Carbon, are you seriously over thinking and over complicating things here?
Am I alone in thinking this?
I just want to understand without having to read your posts five times? Is there anyway to do that without resorting to Y's, MTO's MCust1's etc... I need some kind of legend chart to get through your posts dude!
If we have a 100nit peak white capable projector, based on what we know re the graphs I have submitted previously which actually show in a real waveform that MOST of the image is all happening at under 100nits, Why cant we map the HDR gamma from 0 to 100 nits to 0-75 nits on the display using the multiplier, then take the last 25nits and using that for all the headroom above 100nits on the ST2084 curve?
Is that a stupid idea?
Glad you asked. Wave shaping is not intuitive, especially when one cannot depend on stds alone to solve the problem. Brightening the picture for projectors can be done with CR and gamma ctrls ... this has been done historically in 2K SDR due to the lack of bulb brightness. In HDR, if one brightens the picture with the CR ctrl, one can do the same thing by changing the shape of GammaD by making its slope steeper. In HDR, one can also brighten the picture by increasing the PT ctrl which will decrease Gamma (just like we do in 2K SDR when going from 2.4 to 2.2, Gamma2.2 results in a brighter picture than Gamma2.4). To increase PT by wave shaping the GammaD curve, one can shift the vertical portion of the GammaD curve to the left. That's exactly what the PT control does. The MCust1 curve for 100nit rigs needs to end up looking like its been shifted to the left (PT5 --> PT12).
So when you say to compress the 0-100nit picture (MTO + Specular luminance levels) into 75%, then use the remaining 25% for head room ... it seems like this a GammaD solution for the 4 clipping points (e.g. 1000/0.000, 1000/0.005, 1100/0.005, 4000/0.005). The question is "how do we design the brightness of one GammaD's curve to sync up with all 4 clipping pts of the ST.2084 curve? That's the question ... and your solution is compression/headroom. Yes?
Below is the scenario you are trying to solve. We have just 1 JVC GammaD brightness curve and 4 different clip pts for the ST.2084 curve. GammaD has to be designed so both of these curves sync up and display movies with adequate brightness at all luminance levels, based on 4 different clipping points. The GammaD brightness scale shown below is based on a linear "relative" luminance scale. ST.2084 brightness scale is based on a logarithmic "absolute" luminance scale. GammaD's curve will be re-scaled and the resulting extra head room will sync the single gammaD curve with the 4 different clipping points from the ST.2084 curve ... resulting in a great looking HDR picture, no matter which clipping point the source material uses.
1. JVC's GammaD Relative Brightness at Y5%, Y10%, Y15% ... Y95%, Y100%
vs (has to compliment the 4 different clipping points of the ST.2084 curve)
2.1. ST.2084 Brightness at 5nits, 10nits, 15nits, ... 990nits, 995nits, 1000nits/0.0nits.
2.2. ST.2084 Brightness at 5nits, 10nits, 15nits, ... 990nits, 995nits, 1000nits/0.005nits.
2.3. ST.2084 Brightness at 5nits, 10nits, 15nits, ... 1190nits, 1195nits, 1100nits/0.005nits.
2.4. ST.2084 Brightness at 5nits, 10nits, 15nits, ... 3990nits, 3995nits, 4000nits/0.005nits.
Not sure there is enough precision in a relative luminance curve to do this. That's why ST.2084 went with logarithmic scale ... for the precision. If your solution does move forward, the first test for this solution would be to make sure the "eye eviscerating" specular highlights we enjoy today on our 100nit rigs are not compromised. 4K HDR movies like Batman v Superman, Pacific Rim, Sucide Squad and Ghost Buster should be tested. Secondly, chk for compression artifacts and banding. I can certainly test this. Is this what you are thinking?
Edit: I just read Manni01's real/content solution. I'm so fixated at getting better brighter levels at 100nit Yeak, that missing/rolling off content is not on my radar as much. Once the picture gets brighter at MTO levels with the same Specular brightness we have today, then I can focus on this.