AVS Forum banner

Color My TV World

4K views 28 replies 13 participants last post by  Marc Wielage 
#1 · (Edited)


Video guru Joe Kane talks about color in the UHDTV ecosystem, including the various possible color gamuts, color containers, "connectivity" (what he calls the signal path from image creation to the consumer), the color spectra of various light sources and display types, why BT.2020 isn't the best choice for a display gamut, metamerism, the importance of backward compatibility and why current UHDTVs and even HDTVs needn't become obsolete, answers to chat-room questions, and more.



Like AVS Forum on Facebook
Follow AVS Forum on Twitter
+1 AVS Forum on Google+
 
See less See more
1 1
#2 ·
Scott, that was a fine thought provoking interview. It would have been great to see Joe Kane at the shootout to explore the issues he raised here but I will be away. I did have one question about the connectivity issue he raised. Just what is it that will carry all of these different streams to the display, so it can pick out the rec. 709, or P3 or whatever color gamut is available? Given that that has to be a rather large stream, can HDMI 2.0 do it? I would think display port will be able to, but will any 4K Blu-ray that is eventually announced support display port? I would think that the studios may have an issue with that. Will we need yet another HDMI spec to handle this level of "connectivity"? Thanks.
 
#3 ·
I did have one question about the connectivity issue he raised. Just what is it that will carry all of these different streams to the display, so it can pick out the rec. 709, or P3 or whatever color gamut is available? Given that that has to be a rather large stream, can HDMI 2.0 do it?
I believe the TV will tell the source device which native gamut (RGB primary coordinates) it has, and the source device will perform the necessary corrections/conversions. Which means that there's no increase in bandwidth at all.

Video guru Joe Kane talks about color in the UHDTV ecosystem, including the various possible color gamuts, color containers, "connectivity" (what he calls the signal path from image creation to the consumer), the color spectra of various light sources and display types, why BT.2020 isn't the best choice for a display gamut, metamerism, the importance of backward compatibility and why current UHDTVs and even HDTVs needn't become obsolete, answers to chat-room questions, and more.
Very nice one, thanks. For once I've barely anything to disagree with. One thing I don't understand is the following, though:

When using a container/carrier that is large enough (ACES or XYZ), all colors that are visible to the human eye can already be represented by the container, with the usual 3 color system (RGB). So why would Joe want to go for a 6 color system? If the ACES/XYZ container can already hold all visible colors, what advantage would a 6 color system bring in addition to that?
 
#5 ·
Very interesting discussion. My major question regarding this "connectivity" idea is how much storage space / bandwidth does it take to deliver all of that content all the way to the consumer? I am thinking it has to be astronomical. I understand that it would be downconverted to whatever display technology the consumer has just before the display device, but it still would need to be delivered into the home of the consumer. Currently that delivery method would need to be something like blu-ray disc, streaming, hard drives or backup tapes (LTO, etc). I think hot swap hard drives or possibly backup tapes are the only storage medium that would be viable. The cheapest 1 TB hard drive I could find on Amazon with a quick search is about $55, and 2 TB is about $82. By the time this would come to the consumer, it might be $25 per hard drive. Streaming could work with this by requesting the best signal the display is capable of with the available bandwidth, but it would require massive processing at the carrier's facilities to handle all of the various consumers specific needs.

Another question would be regarding all of the different manufacturers of the capture devices. They would all most likely have different capabilities and would store there content in their "native" format for that capture device. Dealing with all of the various capture devices and making a standard to handle all of them that they can store to and be interpreted by a single playback device on the consumer side might be difficult as well.

I really do like the concept and I get that it makes the sky the limit. I also wonder about going the other way where you have a much better display than the content is stored with (like rec709). I suppose it would be up to the display to upconvert the content (color, etc.) or limit the color so it looks like rec709 on the display.

Mike
 
#7 ·
Very interesting video. Ever sense CES 2014 I've been fascinated by wider color gamut. It's really the next big thing that the TV business needs. Our current UHD is great but the fact is that there isn't much of a difference when sittings at a normal distance. Wider color is something you can see from any distance. However, this video has brought up some great points about the possible issues of Rec.2020.

Perhaps jumping to Rec.2020 is not the best option. IMO bringing P3 to consumer TVs with content to take advantage of it would be a great upgrade. Sure id like something higher than P3 such as the Adobe RBG, but at this point anything is better than the ancient Rec.709 that we are limited to now.

One thing is for sure, CES 2015 will be very interesting. With the disagreement in color standards, the development of HDR and the issues with 4K Blu-Ray, this UHD system seems to be off to a rocky start. We have the technology but we need the displays and content.
 
#8 ·
I am not surprised that Joe Kane believes that DCI P3 would be best for UHDTV since that is what the studios are promoting. While some people can notice a difference with a pure spectral color it is a very small difference (DCI P3 has a pure spectral red and no one complains about it).

The studios are doing their best to kill the Rec. 2020 color space since if it is chosen for UHDTV than the studios would have to upgrade their equipment (and that equipment costs a lot). While the Rec. 2020 space is large there are several display technologies (laser projection, LCD, and OLED) that are capable of showing over 90% of it. For example Nanosys made an LCD display that could show 91% of the Rec. 2020 color space and if it had better color filters it could have shown 97%.
 
#10 ·
I think Rec2020 is total overkill for a lot of stuff, particularly in that very few cameras can capture this standard, and nobody is set up to transfer old films in this standard. The joke in the mastering business is that they call it "Rec2020" because 2020 is probably the year all the kinks will be worked out of the standard.

I'd be happy if people could just see mildly-compressed HD images in proper Rec709 at home. That actually looks pretty good, though I'd agree with Joe that it could stand improving... someday. P3 is very doable, but it's not gigantically and unsubtly different from Rec709:



As you can see in the diagram, getting really strong greens is really hard. The problem with trying to reproduce P3 is that most of that work today is being done with projectors, and there's a lot of reasons why P3 D-Cinema work doesn't work well with monitors. You get an entirely different perception seeing a large 20' image on a projection screen than you do on (say) a 50-60" monitor. It's a totally different experience.

There's a lot of good stuff on color space and color management in general in this free SIGGRAPH tech paper, if anybody wants to read it:

https://raw.githubusercontent.com/jeremyselan/cinematiccolor/master/ves/Cinematic_Color_VES.pdf

Be warned there's a lotta math and a lot of photos with circles and arrows on the back of each one, but the essential science is covered in great detail. This is not a simple subject, and trying to cram this technology in a $695 consumer monitor is fraught with problems.
 
#11 ·
why not simply give them free access to multiply colorspaces.
for example P3, 709 and 2020. the playback device has to be able to transform them all down to 709 so compatibility is given or even better it has to be able to transform all to all supported colorspaces. studios cna now use the best fitting colorspace. if you encode a source with only 709 transform to 2020 the encoder can't use the full colordeep because a lot possibilities are never used so encoding it as 709 should result in better quality.

forcing bt 2020 doesn't sound like a good choice to me. you need more space to store those informations that can't be displays in the next ~10 years and high bit deep is a must have.

it is known that high bit deep can help to compress lower bit deep sources better but in this case all information stored are fully used as intended later on and needed.
 
#12 · (Edited)
quote=eclipsegt;26252313]Very interesting video. Ever sense CES 2014 I've been fascinated by wider color gamut. It's really the next big thing that the TV business needs. Our current UHD is great but the fact is that there isn't much of a difference when sittings at a normal distance. Wider color is something you can see from any distance. However, this video has brought up some great points about the possible issues of Rec.2020.

Perhaps jumping to Rec.2020 is not the best option. IMO bringing P3 to consumer TVs with content to take advantage of it would be a great upgrade. Sure id like something higher than P3 such as the Adobe RBG, but at this point anything is better than the ancient Rec.709 that we are limited to now.

One thing is for sure, CES 2015 will be very interesting. With the disagreement in color standards, the development of HDR and the issues with 4K Blu-Ray, this UHD system seems to be off to a rocky start. We have the technology but we need the displays and content.[/quote]

How can you state and believe OUR current UHD is great? As established by the industry's trade association, CEA, in September a set qualifies to be called a UHD set if it displays 3840 x 2160, it can upscale anything to that, it has an aspect ratio of 1.78 or higher, it can handle 8 bits and it does rec 709. Has to be able to input and display 3840 x 2160 at 24, 30 and 60 fps and have at least one HDMI input. And it must be protected with HDCP 2.2 or its equivalent. HD with more pixels. That's really great.

Will the studios allow P3 for the home? They want to give commercial theaters something better than they give home consumers. My guess is that the coming 4K blu ray standard will be a space not as big as DCI. But the consumer really doesn't care. I can not remember anyone ever posting that they saw a movie at a DCI theater and then watched it at home on blu ray and they noticed less colors at home.

If you can't understand why rec 2020 is bad, watch the video again. The bands are way too narrow and when color is generated with narrow RGB a broad cross section of people will see the colors quite differently from their neighbor.

P3 would be great and would make post work conversion to a 4K blu ray standard that adopted P3 a lot easier and much more cheaper.

Remember when the 4K blu ray standard is announced, and it will be announced I think no later than CES, it will be a political decision and not what idealist want.
 
#17 · (Edited)
Of course for 4K Blu-Ray / UHD we would use h264/HEVC with 16bit integer compression, instead of lossless compression. That should fully remove any bandwidth problems.

ACES uses half precision floating point values which are a lot harder to compress than integer values. ACES has a fantastic dynamic range and very high precision but it is also very expensive.
There's no problem converting floating point integer. It should be easy to use ACES, but store the data in 16bit integer instead of floating point. Actually 16bit integer has more precision than half precision floating point, and HEVC already supports 16bit integer, and it compresses with very good efficiency. Ateme has published some papers which (somewhat surprisingly) show that increasing the encoding bitdepth helps compression efficiency with h264 instead of harming it. Meaning encoding 16bit integer should need ever so slightly less bitrate than encoding 8bit.
 
#18 ·
if you encode a 8 bit rec 709 in 10 bit it is normally better compressed true. but things change here if you take a rec 709/p3 source and change it to ACES and store it in 16 bit some bits are never ever used.
ACES is so much bigger that over 50% of the possibilities in the 16 bit are never used most likely even more. so HEVC has to encode 16 bit data that doesn't make use of the full 16 bit the bit for the highes number is never ever used because the colorpsace is so much bigger. a 8 bit source encoded in 10 or even higher bit deep makes full use of extra bits.
 
#20 ·
i'm pretty sure bt 709 source encoded in 16 bit is better compressed as a bt 709 source encoded in 16 bit in ACES. don't forget the informations are 50 % or even more important in ACES else in bt 709 and i'm pretty sure the HEVC encoder doesn't know that.
and this alone is a huge deal. 10 bit decoding is already a lot slower than 8 bit decoding 16 bit will be really hard and ASIC going to be expensive a no go for something for the masses.

and it's know that 8 bit source are better compressed in 10 bit.
if this is still the case with 16 bit is not 100 % clear yet. and the "trick in 10 bit encoding is that first the lower bits are changed because the lower bit information aren't important but with a bigger colorsapce they get a lot more important so it does the opposite of what high bit deep does to 8 bit sources.

just theoretical and this is worth testing i my opinion. and we aren't missing a lot of tools to test this "soon".
 
#23 · (Edited)
i'm pretty sure bt 709 source encoded in 16 bit is better compressed as a bt 709 source encoded in 16 bit in ACES. don't forget the informations are 50 % or even more important in ACES else in bt 709 and i'm pretty sure the HEVC encoder doesn't know that.
I believe it will make no difference. The encoding efficiency improvements of using a higher bitdepth source comes from having lower rounding errors inside the encoder math, and from the content having finer gradations which makes the image easier to "understand" for the encoder. Sure, when using a large color space, you might lose a bit of precision. So maybe a 16bit ACES source will compress similar to how a 14-15bit BT.709 source would compress. Which is a very very small difference.

10 bit decoding is already a lot slower than 8 bit decoding 16 bit will be really hard and ASIC going to be expensive a no go for something for the masses.
I disagree. 8bit was easy. 10bit is not a size which is often used in the computer world. 16bit is a much more natural number for computers. 8bit = 1byte. 16bit = 2byte. 10bit = 1.25byte. 10bit is a really odd number, as is 12bit. I don't think making 16bit ASICs is much more difficult or expensive than making 10bit ASICs. Of course anything more than 8bit is more expensive than 8bit. But I think 16bit will only be marginally more difficult than 10bit.

and it's know that 8 bit source are better compressed in 10 bit.
if this is still the case with 16 bit is not 100 % clear yet.
Yes, it is. As I said, Ateme has written papers about this.

From I have read, ACES or XYZ is not a colour space. Its a postproduction carrier and it uses metadata, which are essential for "showing" the image in Rec.709, P3, whatever. It covers whole CIE diagram, so that any colour space can be derived from it. If 6 colors would be in such standard, no problem.
http://en.wikipedia.org/wiki/CIE_1931_color_space

"[...] the CIE 1931 RGB color space and CIE 1931 XYZ color space are the first mathematically defined color spaces."

In any case, I never said that using 6 colors would be a problem. I just don't understand the benefits of using 6 colors (for encoding / transport).

I have used 16-bit as a option because simple logic tells me, that if Rec.709 needs at least 10-12bits to eliminate banding and brings things at least where they should be, 12-bit would be superb for P3. Rec.2020 may need 16-bits because now we would need to quantize a much more color information.
FWIW, people often forget that the banding which is most visible to our eyes is luma banding. And the luma channel doesn't care about the source gamut. Whether we use BT.709 or BT.2020, the luma channel still starts at black and ends at white. With 12bit there should be zero visible (luma) banding left. Regardless of the color space. The situation may be different if you compress RGB instead of YCbCr, though. Which is why I'd like to see YCbCr compression to be used - unless we get 16bit, then it shouldn't matter.
 
#22 ·
According to this document about ACES:
https://www.oscars.org/science-technology/council/projects/pdf/ACESOverview.pdf
they use RGB floating point for digital cameras and 16-bit for material from the scanner.

Regarding HEVC, I would be very interested to see such calculation for 12 or 16-bit / 4:4:4 / 3840 x 2160 24fps before compression and after compression.
I have used 16-bit as a option because simple logic tells me, that if Rec.709 needs at least 10-12bits to eliminate banding and brings things at least where they should be, 12-bit would be superb for P3. Rec.2020 may need 16-bits because now we would need to quantize a much more color information. Im writing this because using 12-bits vs using 16-bits 4:4:4 in 2160p matematically should make huge difference in terms of required connectivity for a single cable.
 
#24 ·
yep.By the time this would come to the consumer, it might be $25 per hard drive. Streaming could work with this by requesting the best signal the display is capable of with the available bandwidth, but it would require massive processing at the carrier's facilities to handle all of the various consumers specific needs.
 
#26 ·
AFAIK, the July 2014 HEVC meeting finalized a lot of things they were working on, including full 16bit support etc. From what I've read, HEVC compresses 10bit better than 8bit, too, same as h264, but this is just what I read "somewhere on the internet". The Ateme papers are about h264.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top