HDR Discussion Thread

[SeeMoreDigital]

Originally Posted by Geoff D

Because the 16:9 frame size for consumer video is what it is, and that doesn't just apply to 4K UHD but HD video as well. Adding in the need for players/TVs to generate the borders themselves could've created other issues with encoding/playback and it's far simpler to just hard encode the borders into the image. That said, anyone who plays back video streams which have had the borders cropped out will know that this usually works pretty well in the player/TV, that they just recreate the borders and take it from there.

Indeed. When support for MPEG-4 Part-2 playback started to appear in hardware playback devices (many years ago), most were able to electronically generate black mattes.

I can understand why adding provision for electronically produced/generated black mattes was not included in the hybrid analogue/digital hardware player days but we're all living in an 'all digital' 4K content device world now.

Cheers

[Geoff D]

It was just easier to carry over that way of thinking, seeing as 4K UHD Blu-ray was built atop the bones of Blu-ray and they sought to actually remove features from the BD spec that were never really used all that much. Programming in the gubbins necessary for border generation, no matter how simple it may have been, wasn't gonna happen. They certainly wouldn't have known when debating the spec five years ago that DV dynamic metadata would cause this specific issue because the hardware is doing something stupid. As it is, it seems to be a rather obvious fix: tell the ****ing Dolby Display Manager to stop applying metadata onto the black bars!

[DisplayCalNoob]

Originally Posted by Geoff D

I'm not Penton, but didn't this happen several months ago? There are actually 21 trim pass features, the newly added Mid Tone Offset is not grouped with the Level 2 Primary (8) and Secondary hue & saturation (12) trim controls.

This will probably get lost very quickly but I think I've found the cause of the Dolby greybars problem, this is from the V3.0 DV Best Practices document regarding the L2 trim metadata:

Yeah, I noticed that yesterday. The solution appears to be pretty simple. One of the my reasons for asking v.4.0 trims, is the fact that many of first DV releases were v.2.9 with tone mapping, etc. V.4.0 color specific saturation and hue upgrade, is pretty exciting.

[Geoff D]

It is? I mean, they've added those Secondary trim controls for a reason because colour can, according to the mixinglight.com tutorial, look a little oversaturated when being mapped so now they can fine tune it further, but that's kinda the point: these will be small gains, I doubt that we're talking night and day differences here. It's still just trim metadata, not a wholesale change to the actual grading of the source content which has already taken place by the time the process has got to this stage. And the better the TV is at handling HDR to begin with, the more these mapping gains get even more minute.

[DisplayCalNoob]

I think we are much more likely to see many of those features on streamed DV content first, the more recently released movies could see some artistic mods.

[Geoff D]

But how will you even know what "feature" it is you're seeing? Even with the six former controls, were you sitting there watching DV on whatever TV you have thinking "hmmm, I think they've upped the Chroma Weight Offset there in that shot" or whatever? And, again, this is NOT about adjusting the "artistic" direction of the content for the sake of it once it's been established in the master grade, it's about adding more finesse to the trim (tone mapping) controls. That's it. A noble intent then to be sure, but not likely to result in any visual fireworks in and of itself.

[Penton-Man]

Originally Posted by DisplayCalNoob

Yeah, I noticed that yesterday. The solution appears to be pretty simple. One of the my reasons for asking v.4.0 trims, is the fact that many of first DV releases were v.2.9 with tone mapping, etc. V.4.0 color specific saturation and hue upgrade, is pretty exciting.

more precision with v4.0 –

[DisplayCalNoob]

Originally Posted by Geoff D

It is? I mean, they've added those Secondary trim controls for a reason because colour can, according to the mixinglight.com tutorial, look a little oversaturated when being mapped so now they can fine tune it further, but that's kinda the point: these will be small gains, I doubt that we're talking night and day differences here. It's still just trim metadata, not a wholesale change to the actual grading of the source content which has already taken place by the time the process has got to this stage. And the better the TV is at handling HDR to begin with, the more these mapping gains get even more minute.

I always thought, that the point, giving the artistic control after the metadata was generated. The colorist due to studio request, could adjust hue and saturation of emergency vehicle lights as an example, so that they cast and reflect the way they should.

When I watched most of the HDR content I have early on, I thought it looked great until I calibrated Gain to D65, I relalized how image information was being clipped. When I adjusted hue and saturation, for all 6 colors, reflections, and more detail in the image became more apparently clear.

So proper color saturation and hue, means closer to the HVS, at least to me it does.

[Penton-Man]

anecdotal, superficial comparison of various HDR outcomes - https://www.youtube.com/watch?v=djuEF6tQZfc#t=6m55s

shout out to Vincent T. and/or any eventual 2019 Sony TV owners, evaluate HDR10 vs. HDR10+ as proposed here - https://forum.blu-ray.com/showthread...e#post15975642

[Penton-Man]

recent DV facility certification news….

[Penton-Man]

Reminder to Robert Z. and other New Yorkers - https://www.eventbrite.com/e/smpte-n...on-53830601779
I’m interested in knowing if Mark shows up in shorts. I dare him to.

[Robert Zohn]

I'm registered and have a few folks I know well on the panel. Looking forward to this SMPTE NY special discussion event.

[Geoff D]

Originally Posted by DisplayCalNoob

I always thought, that the point, giving the artistic control after the metadata was generated. The colorist due to studio request, could adjust hue and saturation of emergency vehicle lights as an example, so that they cast and reflect the way they should.

When I watched most of the HDR content I have early on, I thought it looked great until I calibrated Gain to D65, I relalized how image information was being clipped. When I adjusted hue and saturation, for all 6 colors, reflections, and more detail in the image became more apparently clear.

So proper color saturation and hue, means closer to the HVS, at least to me it does.

No. The grade is the grade is the grade, the colourists already have access to a vast multitude of controls that can tweak the image any way they (and hopefully the filmmaking talent) see fit. Once that's signed off, then the metadata creation is undertaken. What you're referencing here are the TRIM controls that are used to alter the downconversion metadata when mapped into specific circumstances e.g. SDR 709 100-nit, HDR 600 nit, HDR 1000 nit etc, NOT controls that are intended to entirely alter the creative direction of the image after the fact due to "studio request".

[Penton-Man]

Originally Posted by Robert Zohn

I'm registered

one downside to having attended CES and taking photos….https://www.bbc.com/news/technology-46875947
pass that on to Larry

[Penton-Man]

In other words….

Originally Posted by Geoff D

The grade is

the HDR master sequence encoded and exported as 16 Bit Tiff or EXR files. The files are created with the PQ curve as part of the rendered images.

And then afterwards another component produced during a Dolby Vision grading session is the Dolby Vision XML file that contains metadata about the project, the image analysis and trim data for every additional target display. This XML is used to create different masters and also dynamically map the content on Dolby Vision enabled displays based on the additional target display trim data.

Without getting into who has control/authority over various stages of the process and what happens when the Director or DP is not in the room, you’re both kind of right, Geoff more so though as he understands the process better. As Geoff says “What you're referencing here are the TRIM controls that are used to alter the downconversion metadata when mapped into specific circumstances e.g. SDR 709 100-nit, HDR 600 nit, HDR 1000 nit etc, NOT controls that are intended to entirely alter the creative direction of the image after the fact due to "studio request".

On the other hand, at the end of the day I think DisplayCalNoob wants the final image on a person's TV at home to “cast and reflect the way they should….. So proper color saturation and hue, means closer to the HVS, at least to me it does”. With that goal in mind, the added precision afforded in v4.0 thru the primary and secondaries is a technological attempt to allow the mastering operator to make the movie on the target display (Geoff's "specific circumstances") look as close as possible to creative intent/that seen on the grading monitor.

[Penton-Man]

Originally Posted by Staying Salty

BDA Talks 4K Ultra HD Blu-ray, 8K TV & Streaming at CES 2019 - YouTube

The BDA ^, the CTA, the Ultra HD Forum, the Ultra HD Alliance….time to add a couple more alliances/associations acronyms to the AV world -

SVA – https://www.streamingvideoalliance.o...ideo-alliance/
8KA - https://8kassociation.com/

[Penton-Man]

Originally Posted by DisplayCalNoob

the HVS,

And what it’s worth - https://www.nbcnews.com/health/healt...50-000-n834261
ya’ll with excellent vision, consider yourselves gifted and take care of dem eyes

[DisplayCalNoob]

Originally Posted by Geoff D

No. The grade is the grade is the grade, the colourists already have access to a vast multitude of controls that can tweak the image any way they (and hopefully the filmmaking talent) see fit. Once that's signed off, then the metadata creation is undertaken. What you're referencing here are the TRIM controls that are used to alter the downconversion metadata when mapped into specific circumstances e.g. SDR 709 100-nit, HDR 600 nit, HDR 1000 nit etc, NOT controls that are intended to entirely alter the creative direction of the image after the fact due to "studio request".

I got it. Wish they would do the higher nit trim pass anyway. The dynamic metadata/tone mapping does a pretty good job of handling brightness and color saturation, on most displays. There are just some parts of the PQ, I wonder, still need that trim pass, especially when it comes to vehicle tail lights, that appear to be over saturated(red). Since there is no reason, why the detail in vehicle tail lights should be lost. Unless the HDR grade has it to, which would be even more disappointing.

[Staying Salty]

https://www.avsforum.com/forum/40-ol...l#post57446288

Originally Posted by Mark Rejhon

Ironically about 3D...
Whether for 3D lovers or haters -- history note -- it's worth noting, that it was 3D that led to CRT-motion-clarity on desktop gaming LCD monitors (e.g. LightBoost, ULMB, etc). Those had crappy uniformity, colors, contrast, but shook traditional notions about LCD motion clarity. LightBoost was originally designed for 3D glasses but apparently became more popular as a motion blur reduction mode. Blur reduction survived while 3D did not, and many gaming monitors still have the blur reduction modes long after discontinuing 3D modes.

The 120Hz+BFI OLEDs are likely one of the world's best 3D televisions (less 3D crosstalk) -- with third party appliances/glasses, even though the televisions are no longer designed for 3D and the manufacturers never tested them for 3D compatibility.

It is the monitors with good blur-reduction that are very good "hackable to 3D" monitors. Meaning, we've (and others) succeeded in getting excellent 3D from displays normally not designed for 3D glasses -- as long as the display had an excellent motion blur reduction mode during 120Hz.

Also, from an engineering perspective... Ultra-Hz also helps improve color depth for HDR algorithms, e.g. by adding temporal dithering (ultra-bitdepth FRC algorithms) to convert 8-bit/10-bit into 12-bit or even 14-bit, even for lower refresh rates, without adding visible temporal dither artifacts (thanks to ultra-Hz). Some of the insanity of bitdepth is necessary for potential future anti-vignette/anti-band algorithms. Preventing low-Hz temporal dithering that is annoying to eyes. There are some great applications of ultra-Hz that can be applied to improving HDR color quality and dim-image-quality of low-Hz.

With some auto-calibration (e.g. feedbacking high-def calibration camera into the OLED electronics -- onetime or occasional use (ala ISF calibrator) -- or embedding light/pixel-leakage monitoring within the panel / behind the panel) it may even be possible to self-capture the darkfields of OLED to do anti-band/anti-vignette compensatory imagedata (a realtime ultra-Hz temporally dithered inverse image that cancels-out, at ultraprecise bitdepths made possible via ultra-Hz) that makes every fractional IRE of a darkfield of OLED perfectly solid uniform with zero visible temporal dither artifacts, zero banding, zero vignetting!.

So, that's why in applied display engineering, ultra-Hz has a huge potential, algorithmically, to fix backplane uniformity problems, too.

Also, keep in mind temporal dither at 10-bit converting to 14-bit or even more is ultra-faint compared to 1-bit DLP temporal dither pixel flicker. You're only dealing with alternating shades of colors only 1/1024 apart to generate the extra bits. Now, consider that this will be achieved via ultra-Hz too, so you have none of the low-frequency FRC flicker at all. Normally these extra bits are way excess overkill but are needed at the dark end for high quality realtime anti-vignette/anti-band algorithms in future OLED panels via realtime inverse-images that perfectly balance-out the banding/vignette realtime custom generated automatically on a per-panel basis.

(Real World Textbook Case Study Equivalent: SLR darkfields that de-noises long camera exposures. Same principle. This post describes the OLED equivalent, which is massively more complex but much easier/simplified with ultra-Hz)

So there's eye candy for low-Hz use cases too!

This is above and beyond knowing that the technology requirements (of 3D, and of blur reduction) already have a very heavily-overlapping venn diagram.

Seems like some potential goodies for everyone.

[Geoff D]

Originally Posted by DisplayCalNoob

I got it. Wish they would do the higher nit trim pass anyway. The dynamic metadata/tone mapping does a pretty good job of handling brightness and color saturation, on most displays. There are just some parts of the PQ, I wonder, still need that trim pass, especially when it comes to vehicle tail lights, that appear to be over saturated(red). Since there is no reason, why the detail in vehicle tail lights should be lost. Unless the HDR grade has it to, which would be even more disappointing.

Does everything have to be brought back though? I'm not saying this to be argumentative, I'm just aksing a question. Just because HDR can do it doesn't mean it should do it, and if someone grades a tail light to look blown out then I'm happy enough with that decision. I was reminded of a similar thing just now when reading back my comments about the disc in the TMNT UHD thread, that the tail lights which skewed to a hot pink in the SDR grade do exactly the same thing in the HDR grade. [edit] As much as I praise the naturalism that HDR imparts onto images - and I do it a lot - there's also room for more stylised, exaggerated visual interpretations that don't always conform to what the HVS can/will accommodate; this is still make believe after all, and not real life.

Still, you raising the point about the extra trim controls does bring to mind what happened with Cuaron's Roma, that when they delivered the Dolby Vision version to Netflix they also delivered a specifically graded SDR 709 version - apparently the first time this has happened - because Cuaron was not happy with how the downconverted SDR looked. One wonders if they only had v2.9 available while grading.