Real-world support for full range and greater color depths

Originally Posted by benwaggoner

So how does it work then? The extended primaries get truncated or something? Or is it some kind of an enhancement layer? Multiple encodes?

I don't think anyone has analyzed a stream yet to know exactly. I assume it's some form of custom signaling or metadata that their players can read and is ignored by others. There's no way it can be multiple encodes when the other point of these discs are to increase the average bitrate.

To add from here:

Other Blu-rays have been sourced from a 4K master, so what’s the big deal with these? A few things. First, the video bitrate has been upped considerably to ensure a solid, artifact-free picture: The new, Mastered in 4K Spider-Man maintains a steady 35 Mbps, as compared with the high-20 Mbps bitrate typical with other discs, including the 2012 Spider-Man Blu-ray. (Unfortunately, the increased video data rate means there’s no space for extra stuff like features and commentary tracks.) Second, Mastered in 4K discs incorporate something Sony calls Expanded Color, which otherwise goes by the name x.v.Color or xvYCC. To briefly sum up, discs with x.v.Color incorporate data describing colors outside the standard Rec.709 HDTV color space. When a Mastered in 4K disc is played on an x.v.Color-compatible BD player (according to Sony, only its own player line plus the PlayStation 3 can reliably be called compatible) and displayed on a TV capable of x.v.Color reproduction — Sony’s new W900A series HDTVs (look for my review in the June/July/August issue of Sound&Vision) and XBR-X900A Series 4K Ultra HD TVs, for example — you get the option to view this expanded range. And displaying it shouldn't involve any shift in the TV’s color points — the disc simply instructs the TV to “retrieve” those extra colors without distorting the set’s overall color balance.

There's not a whole lot of specifics though from what I've seen.

[benwaggoner]

Originally Posted by paradoxical

To add from There's not a whole lot of specifics though from what I've seen.

There really aren't.

The only obvious way I see for them to do what they're doing would be to have a stock Rec. 709 encode and then some kind of enhancement later that added the (relatively small delta) of xvYCC data. Since it's mostly in chroma, that wouldn't be that many samples and they can be quantized quite coarsely. Perhaps using something like Scalable Video coding. I'd think the PS3 would likely have the horsepower to do something like this.

I am now very intrigued by this!

Also, other than Sony BD players, are there any known devices capable of xvYCC playout? It seems support is way more common in TV than it is in any way to play back the content. Can decoders internal to the TVs do the right thing with xvYCC bitstream?

[IanB]

From what I understand Sony use the 1-15 and 241-254 chroma values to represent higher saturation colours. When you do the normal YUV->RGB translation you can end up with normalised colour values less then 0.0 or greater than 1.0, normal decoders clamp these values, advanced decoders map these values into a wider gamut colour space. The luma range is still 16-235.

[benwaggoner]

Originally Posted by IanB

From what I understand Sony use the 1-15 and 241-254 chroma values to represent higher saturation colours. When you do the normal YUV->RGB translation you can end up with normalised colour values less then 0.0 or greater than 1.0, normal decoders clamp these values, advanced decoders map these values into a wider gamut colour space. The luma range is still 16-235.

That makes sense. But if chroma is getting clamped at those values, I worry we'll see posterization of chroma levels when the extended colors get clamped. A gradient of "very red" could turn into a flat area of just red without any detail. If there is some texture in the luma channel that might mask things, but still. It seems like it would be hard to make a "best of both worlds" encode like this.

[IanB]

The out of gamut high saturation colours considered here would have been clamped in normal processing anyway, either by the source camera and/or the display device. The system just gives the opportunity to not clamp the very saturated colours when a high gamut source and display are available.

[benwaggoner]

Originally Posted by IanB

The out of gamut high saturation colours considered here would have been clamped in normal processing anyway, either by the source camera and/or the display device. The system just gives the opportunity to not clamp the very saturated colours when a high gamut source and display are available.

But if you allow out of range colors that will get clamped on some devices, it seems inevitable that you'll get posterization when it does wind up getting clamped.

[IanB]

Yes, intuitively I agree there should be some sort of saturation posterization effect but how do you characterise and witness it.

We are talking about saturation getting clamped here, not luminance or hue. And saturation restriction happens with our simple trichromatic eyes.

The clamping happens in the YUV to RGB translation where you end up with normalised RGB values less then 0.0 or greater than 1.0 I tried an experiment using Limiter(0, 255, 31, 226) to knock another 15 off the normal chroma range and I cannot see any difference, which just probably means I do not have samples that sufficiently use the high saturation part of the existing gamut and/or a display that is not already more saturation challenged than the clamp.

I guess an extreme test case might involve some pure monochromatic light sources like sodium lamps, lasers, etc and a display version of the same. But our eyes cannot see spectrally pure colours as spectrally pure.

[benwaggoner]

Originally Posted by IanB

The clamping happens in the YUV to RGB translation where you end up with normalised RGB values less then 0.0 or greater than 1.0 I tried an experiment using Limiter(0, 255, 31, 226) to knock another 15 off the normal chroma range and I cannot see any difference, which just probably means I do not have samples that sufficiently use the high saturation part of the existing gamut and/or a display that is not already more saturation challenged than the clamp.

Yeah. Begging the question of whether xvYCC actually provides any viewer value for most real-world content. How different is the master, really, and for what content?

[ChiDragon]

When I compared the new "Mastered in 4K" BD of the 2002 Spider-Man movie to the old disc, I found that there were more instances of >240 chroma on the previous release than the xvYCC one.

It would be nice if they included some test content to verify the effects of the gamut expansion on your TV. There is a hidden video file but it's just a long, slow pan across a still image with no apparent usefulness.

[benwaggoner]

Originally Posted by ChiDragon

When I compared the new "Mastered in 4K" BD of the 2002 Spider-Man movie to the old disc, I found that there were more instances of >240 chroma on the previous release than the xvYCC one.

Huh.

Are there any discs that are known for particularly good use of xvYCC? Or even noticeable use?

[SeeMoreDigital]

Originally Posted by benwaggoner

Huh.

Are there any discs that are known for particularly good use of xvYCC? Or even noticeable use?

I'd go further...

I would like to know if there's anybody on this forum who has a 'xvYCC supporting' Sony Blu-ray player and TV? Because without these two supporting devices, nobody is going to 'see' any noticeable difference!

[benwaggoner]

Well, I've made myself a Rec. 2020 H.264 High 10 encode via x264!

Anyone have any idea what players actually know about Rec. 2020? I've got a Quadro K4000 plugged into a Dell U3011 30-bit display via DisplayPort, so I can render out 10-bit images via both DirectX and OpenGL. Dell claims it has a 120% gamut range based on the CIE1976 (83%) and CIE1931 (72%) test standards.

But what should be able to do the right thing here?

[vivan]

madVR

[mp3dom]

Since the topic also mentions greater color depth, I would like to point out the Panasonic "Master Grade Video Coding". There are not a lot of infos about it, but seems that the hint is to use the bluray MVC structure to use the base view as the standard image, and the dependent view for increase the bit depth of the video. Panasonic players with MGVC support should be able to 'merge' the MVC and outputs a 2D video with up to 12bit informations. Also, this method allows to fully use the 60 Mbps of the bluray 3D specs.
Here some explanation (JP)

[sneaker_ger]

Originally Posted by vivan

madVR

madVR does not support output higher than 8 bit - just input. (Plus it only outputs RGB).

[vivan]

Originally Posted by sneaker_ger

madVR does not support output higher than 8 bit - just input. (Plus it only outputs RGB).

But it does support Rec. 2020. I doubt that any other renderer supports it. So it's better than nothing, right?

[benwaggoner]

Also, what's the proper --colormatrix setting for 10-bit Rec. 2020 (without expanded luma range)?

Using --colorprim bt2020 --transfer bt2020-10 are obvious, but I can't find any documentation on the difference between bt2020c and bt2020nc

[drmpeg]

Originally Posted by benwaggoner

Also, what's the proper --colormatrix setting for 10-bit Rec. 2020 (without expanded luma range)?

Using --colorprim bt2020 --transfer bt2020-10 are obvious, but I can't find any documentation on the difference between bt2020c and bt2020nc

bt2020c = constant luminance

bt2020nc = non-constant luminance

So for your example, bt2020nc would be correct.

Ron

[benwaggoner]

Originally Posted by drmpeg

bt2020c = constant luminance

bt2020nc = non-constant luminance

So for your example, bt2020nc would be correct.

Why? I would have thought that "non-constant" would refer to the 12-bit mode with high dynamic range luma, which has a different, steeper gamma for high values.