4k h.265 test patterns in rec.2020 color space - Page 2 - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
 4Likes
Reply
 
Thread Tools
post #31 of 55 Old 09-14-2014, 04:38 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,034
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 85
Quote:
Originally Posted by randal_r View Post
I feel it is safe to summarize that we agree that displays of today can not display a REC-2020 color space. Until proven incorrect, there are no test patterns in a REC-2020 color space (in any pixel density format), available to us the public at this time.
I think you might be misunderstanding something here. Test patterns don't generally have a color space - they just instruct the display what values to use for each channel in each pixel. If you happen to have a wide gamut display, then the test pattern will render accordingly.

Quote:
Originally Posted by randal_r View Post
I can not agree with Zoyd where he says, "yes, no need to remaster legacy material." To move legacy material to newer technologies a conversion or remastering process is always required.
No, you don't need to remaster the content. Rec 709 content should be viewed within the Rec 709 color gamut, and a good wide gamut monitor, whether it be Adobe RGB or Rec 2020, will be able to "shrink" its gamut so Rec 709 content is rendered faithfully.
spacediver is offline  
Sponsored Links
Advertisement
 
post #32 of 55 Old 09-15-2014, 07:35 AM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
I have reflected upon your last posting and here is what you have said;

"I said there is no need to, your current HD version of "Gone With The Wind" will be fine when a 2020 display upscales it and emulates 709 primaries (with the caveat I mentioned regarding standard observer functions/narrow primaries). Of course if you want a native 4K version you'll have to re-digitize and remaster using 2020 primaries."

First you presume that the material content is in a HD, BT709 format. There is a mass of material that is in BT601, 480i., how will this material come out? Again, assumptions are that displays will be 4K. Although the ITU-R BT.2020 documentation states the 4K format as minimum, you will find that a more realistic pixel format will be 8K or better. Another common assumption is that the REC-2020 will come out in the year 2020, no where in the ITU-R BT.2020 documentation is this ever mentioned or implied a date to be executed. By some freak event this standard could come into existence next year.

In a prior posting of yours you said;
"yes, no need to remaster legacy material. The problem however is that when the standard observer functions are used to model human vision there is higher real observer color variability as primaries become narrower. This is already surfacing with current technology wide gamut displays and will only get worse."

From the aforementioned statement I fail to see how you were conveying this statement;
"I said there is no need to, your current HD version of "Gone With The Wind" will be fine when a 2020 display upscales it and emulates 709 primaries (with the caveat I mentioned regarding standard observer functions/narrow primaries). Of course if you want a native 4K version you'll have to re-digitize and remaster using 2020 primaries"

I am always curious when someone brings up the subject of "Standard Observer", are they referring to the 1931 Standard Observer (2-degree observer) or the 1964 Observer (10 Degree Observer)? Sometimes I feel that there is to much obsession with this.

To fore go the conversion/remastering process for any of the archived material (in my opinion), would be a mistake. Since people want quality they will not settle for the a display's ability to upscale and produce an adequate image. Many find the the upscaling process of today from a 2K to a 4K (with the same color space Bt709), renders an image that is too soft. Now image the image quality from a 2K or a 480i to an 8K with a completely different color space.

Last edited by randal_r; 09-15-2014 at 08:34 AM.
randal_r is online now  
post #33 of 55 Old 09-15-2014, 07:48 AM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
Quote:
Originally Posted by dr1394 View Post
Geez, I post a nice BT-2020-1 test pattern and this is the thanks I get? Have a heart!

Ron
I do apologize, it may very well be a REC2020 test pattern. In attempting to access it, I was not able to download and verify the results.
randal_r is online now  
 
post #34 of 55 Old 09-15-2014, 08:05 AM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
Quote:
Originally Posted by spacediver View Post
I think you might be misunderstanding something here. Test patterns don't generally have a color space - they just instruct the display what values to use for each channel in each pixel. If you happen to have a wide gamut display, then the test pattern will render accordingly.



No, you don't need to remaster the content. Rec 709 content should be viewed within the Rec 709 color gamut, and a good wide gamut monitor, whether it be Adobe RGB or Rec 2020, will be able to "shrink" its gamut so Rec 709 content is rendered faithfully.

You may be correct regarding test patterns and color space. I am curious about something that you maybe able to shed light upon. Why is there test patterns dedicated to BT601, BT709 and Pal? Each version has different coordinate sets for their primaries, does this not constitute a color space? Can I use a BT601 test pattern to calibrate my HD display? This would have saved me a lot of money to use an older video generator that had only the BT601, instead of buying an Accpel 5000 with the BT709. (The Accupel 5000 is an excellent video generator). Do test patterns for the Adobe RGB color space work for BT709? It is my hope that you will grace me with your wisdom in this matter.


With your regards to your statement, "No, you don't need to remaster the content". Please see posting #32 .


You made a statement in a prior posting;
"in a PC RGB context you could just use any image that has pure primaries (e.g. in an 8 bit context [255 0 0] [0 255 0] [0 0 255]) and so long as your software doesn't do any color management, and allows the display to render natively, you'll render the image at those wide gamut coordinates."

Since the Rec. ITU-R BT.2020 documentation does not allow for 8 bit context and only 10 & 12 bit, it is save to say that having a computer graphics card render REC2020 images is not likely.

Last edited by randal_r; 09-15-2014 at 08:37 AM.
randal_r is online now  
post #35 of 55 Old 09-15-2014, 10:33 AM
AVS Forum Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,936
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 215
Quote:
Originally Posted by randal_r View Post
I have reflected upon your last posting and here is what you have said;
First you presume that the material content is in a HD, BT709 format. There is a mass of material that is in BT601, 480i., how will this material come out? Again, assumptions are that displays will be 4K. Although the ITU-R BT.2020 documentation states the 4K format as minimum, you will find that a more realistic pixel format will be 8K or better. Another common assumption is that the REC-2020 will come out in the year 2020, no where in the ITU-R BT.2020 documentation is this ever mentioned or implied a date to be executed. By some freak event this standard could come into existence next year.
REC601 is another trivial case. REC2020 covers all previous color spaces. All one has to do is 1) use the proper decode matrix and 2) "pull" the display primaries back to the REC601 or REC709 definitions. We already do this with "REC709" displays that have ridiculously "wide" primary color spaces ... assuming said display has a functional CMS. I can't think of any technical reason why a 4K/UHD/REC2020 display would *not* be able to properly display REC601 or REC709. In short, you're getting all worked up over nothing.

That being said, I doubt many people will be "happy" watching 480i/SD material on a 4K+ display ... For the most part, it already looks pretty ugly at 1080p/720p.

Last edited by HDTVChallenged; 09-15-2014 at 10:37 AM.
HDTVChallenged is offline  
post #36 of 55 Old 09-15-2014, 01:08 PM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
Quote:
Originally Posted by HDTVChallenged View Post
REC601 is another trivial case. REC2020 covers all previous color spaces. All one has to do is 1) use the proper decode matrix and 2) "pull" the display primaries back to the REC601 or REC709 definitions. We already do this with "REC709" displays that have ridiculously "wide" primary color spaces ... assuming said display has a functional CMS. I can't think of any technical reason why a 4K/UHD/REC2020 display would *not* be able to properly display REC601 or REC709. In short, you're getting all worked up over nothing.

That being said, I doubt many people will be "happy" watching 480i/SD material on a 4K+ display ... For the most part, it already looks pretty ugly at 1080p/720p.
I thank you for the response. Could you please list all of the documentation that states that the REC2020 covers all color spaces. To date, I have yet to come across such a statement. Since I have never seen a display utilizing the REC2020 format can you tell me at which electronic exhibition you have personally seen a REC2020 display and it displaying a REC709 image?

If a display has a ridiculously "wide" gamut then one can assume that the display is not working within the REC709 parameters and hence the reason to calibrate the display so that it is compliant. You mentioned a "functional CMS", what if the CMS is not functioning? So far in this thread I have heard nothing but opinions being past off as fact. If one were to scrutinize all who had contributed to this thread, I do not think any of them have seen a REC2020 display and if they had, I can almost be assured that the image was not REC709 or less. So how can a statement be made that a REC2020 display will properly display REC601 or REC709?

You misinterpret that I am getting worked up, this is far from the truth. My hope is that all the answers will come to light. In doing so one must avoid assumptions and question all that is placed before them. Emotions to be set aside and information viewed in the harsh light of reality. For example the statement, " I can't think of any technical reason why a 4K/UHD/REC2020 display would *not* be able to properly display REC601 or REC709", do you have a degree in Electronic focusing on video displays or is this merely an opinion? Based on the grammar, it tends to lean towards an opinion.

In your last statement, "That being said, I doubt many people will be "happy" watching 480i/SD material on a 4K+ display ... For the most part, it already looks pretty ugly at 1080p/720p".

Again I see this as an opinion, you like many others who fail to take into account of those whose viewing pleasures lay within another sigma on a demographic chart. I know many who enjoy movies from as early as the 1930's to the 1960's. Are you going to tell them that you feel that their viewing preferences have to be abandoned because you don't think they will be "happy" watching 480i/SD material on a 4K display? Hence the reason I had made statements about the need for conversion/remastering of older and archived material throughout this thread.

With clarity on the subject, I await your reply. I look forward to reviewing the information that you will be forwarding regarding the REC2020 being an all in one color space.

Last edited by randal_r; 09-15-2014 at 01:13 PM.
randal_r is online now  
post #37 of 55 Old 09-15-2014, 05:45 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,034
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 85
Quote:
Originally Posted by randal_r View Post
You may be correct regarding test patterns and color space. I am curious about something that you maybe able to shed light upon. Why is there test patterns dedicated to BT601, BT709 and Pal? Each version has different coordinate sets for their primaries, does this not constitute a color space? Can I use a BT601 test pattern to calibrate my HD display? This would have saved me a lot of money to use an older video generator that had only the BT601, instead of buying an Accpel 5000 with the BT709. (The Accupel 5000 is an excellent video generator). Do test patterns for the Adobe RGB color space work for BT709? It is my hope that you will grace me with your wisdom in this matter.
My comments were restricted purely to gamut considerations, not things like bit depth, resolution, frame rate, or luma coefficients. Let me try to be clearer. If you have an RGB test pattern that instructs the display to only "light up" one channel, then the color that the display renders will depend entirely on the display's native primaries (assuming no color management is in place). If you have an sRGB display and a display that somehow had the rec 2020 primaries, then the resulting image will reflect the gamut of the display.

Quote:
Originally Posted by randal_r View Post
With your regards to your statement, "No, you don't need to remaster the content". Please see posting #32 .
This is your original quote:

Quote:
Other concerns are legacy issues. Do we abandon movies from the past? There is a mass of material that would have to be converted and due to the cost of the conversion much of the material would disappear into oblivion.
You make it seem as if rec2020 displays will somehow make older material obsolete as one would need to convert (remaster) the material. So long as a rec 2020 display can remap its color gamut to Rec 709, then there will be no difference viewing content on that remapped display as there is on current Rec 709 displays (assuming of course that all other Rec 709 specs are able to be emulated on this future display).



Quote:
You made a statement in a prior posting;
"in a PC RGB context you could just use any image that has pure primaries (e.g. in an 8 bit context [255 0 0] [0 255 0] [0 0 255]) and so long as your software doesn't do any color management, and allows the display to render natively, you'll render the image at those wide gamut coordinates."

Since the Rec. ITU-R BT.2020 documentation does not allow for 8 bit context and only 10 & 12 bit, it is save to say that having a computer graphics card render REC2020 images is not likely.
Again, I was only talking about gamut considerations, not resolution or color depth.
Chad B likes this.
spacediver is offline  
post #38 of 55 Old 09-15-2014, 07:47 PM - Thread Starter
AVS Forum Special Member
 
nuke's Avatar
 
Join Date: Jul 2001
Location: CA, USA
Posts: 1,290
Mentioned: 57 Post(s)
Tagged: 0 Thread(s)
Quoted: 242 Post(s)
Liked: 138
Quote:
Originally Posted by dr1394 View Post
Here's something to experiment with.

http://www.w6rz.net/movingbaruhd.mkv

It's H.264 2880x2160 10-bit 75% color bars. I didn't have quite enough memory for 3840x2160, so I used 2880x2160 (which is a valid UHD resolution in the DVB specification).

https://www.dvb.org/resources/public...-avc-mpeg2.pdf

The video bitstream is signalling BT.2020 non-constant luminance parameters and the YCbCr values are calculated using Y' = 0.2627 R' + 0.6780G' + 0.0593B'

Ron
Thanks Ron!!!

I'll play with it as soon as I can figure out how to get either the TV or the Sony 4k player to recognize it.

nuke
nuke is offline  
post #39 of 55 Old 09-15-2014, 08:08 PM - Thread Starter
AVS Forum Special Member
 
nuke's Avatar
 
Join Date: Jul 2001
Location: CA, USA
Posts: 1,290
Mentioned: 57 Post(s)
Tagged: 0 Thread(s)
Quoted: 242 Post(s)
Liked: 138
Ok, so here's the reason for my request.

Here's the native LCD panel gamut of the Sony XBR65X900A 4k UHD display with Quantum Dot backlighting by Color IQ. Even when looking at the spectrum in a simple optical spectrometer, the backlighting is unlike any other set I've ever seen. The idea is that quantum dots emit nice, tight, narrow spectra in green and red while stimulated by blue LED backlight. This gives the LCD a wide gamut by virtue of the low color cross-talk between RGB stripe filters on the LCD panel.

I measured this below by hacking the service menu and using one of the built-in test modes that directly controls the LCD panel. These are the native primaries of the display with no color processing. That's the entire gamut of the display, as much pure red, pure green and pure blue that it can possibly make. I also snagged the 100% saturated secondary colors and the panel's own native white point.

This is the best I can measure with an X-Rite Eye 1 Display Pro 3 (colorimeter). A reference spectroradiometer might have better results, but I don't have $25,000 handy to lay hands on one.

So, this native gamut is pretty wide. Every test source I have is 1080, thus the TV dutifully maps it all into the Rec.709 gamut (smaller black triangle).

I'm curious what happens if I can get some Rec2020 test signals into 4k format, that the attached Sony FMPX10 4k player can play. That is so far, the ONLY way to get a real 4k signal into the TV. And it does "weird stuff" when the Sony 4k box plays into the 4k display. The boxes "talk" to each other and the TV changes modes and things that I can't get it to do when giving it a 4k signal from a computer.

I can get the display go outside the rec 709 gamut by engaging stuff like "live color" and "vivid" mode on 1080p signals. But of course, they just look cartoonish and awful.

So, for grins and giggles, I just want to see what happens if a file constructed of test signals, encoded with all the flag bits that say, "I'm a big boy 4k signal with a 2020 gamut" will scoot outside the 709 gamut and use any of the available gamut at all.

I 'effin love science and all.


Attached Thumbnails
Click image for larger version

Name:	CIE-gamut-native.jpg
Views:	893
Size:	117.3 KB
ID:	264746  

nuke
nuke is offline  
post #40 of 55 Old 09-15-2014, 09:29 PM
AVS Forum Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,385
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 56
I've updated the pattern with slightly more accurate YCbCr values (by using all 10-bits instead of just shifting 8-bit values). Also, since it compresses to a fairly small file, I've extended the play length to 5 minutes.

http://www.w6rz.net/movingbaruhd5.mkv

Here's the raw H.264 bitstream in case it needs to be muxed into another container.

http://www.w6rz.net/movingbaruhd5.264

You may have to right click on the links and select "Save Link As..." to properly download the file.

Here's the source code (that shows the YCbCr values used).

http://www.w6rz.net/movingbaruhd.c

The shift left by 6 is to satisfy the "yuv422p16le" input format used by x264. I'm not sure where that comes from, since the H.264 reference decoder doesn't do that. The x264 command line is:
Code:
x264 -o movingbaruhd.mkv movingbaruhd.yuv --input-res 2880x2160 --input-csp yuv422p16le --demuxer lavf --fps 24 --sar 4:3 --colorprim bt2020 --transfer bt2020-10 --colormatrix bt2020nc
Here's the BT.2020 signalling in the H.264 bitstream.
Code:
   vui_parameters_present_flag: 1
    aspect_ratio_info_present_flag: 1
     aspect_ratio_idc:14
    overscan_info_present_flag: 0
    video_signal_info_present_flag: 1
     video_format: 5
     video_full_range_flag: 0
     colour_description_present_flag: 1
      colour_primaries: 9
      transfer_characteristics: 14
      matrix_coefficients: 9
    chroma_loc_info_present_flag: 0
colour primaries = 9 signals BT.2020 primaries
transfer_characteristics = 14 signals BT.2020 gamma for a 10 bit system
matrix_coefficients = 9 signals BT.2020 non-constant luminance system

For creating test patterns, the only parameter that matters is the matrix coefficients (since these are used to convert RGB to YCbCr). Primaries and gamma are entirely up to the display to handle properly.

Ron

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is online now  
post #41 of 55 Old 09-16-2014, 12:03 AM
AVS Forum Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,936
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 215
Quote:
Originally Posted by randal_r View Post
I thank you for the response.

With clarity on the subject, I await your reply.
You may be waiting for a long time. To quote Mr. Spock, 'To have a meaningful discussion of the subject, one must first have a common frame of reference.'

To point you in the right direction, you might want to try plotting the REC2020 color space (the definitions of primary coordinates in CIE31) along with the corresponding primary coordinates of the REC709 and REC601 color spaces. From this exercise, you *should* find that REC2020 can easily produce *all* of the colors present in REC709 and REC601. Therefore, it is a trivial mathematical problem to map REC709 or REC601 colors into REC2020 space.

It's pointless to "debate" this further ... either you understand the issue, or you need to do a little more study ... and reasoning.

PS: It's just math (linear algebra specifically and a little geometry to help out with the visualization.)

Last edited by HDTVChallenged; 09-16-2014 at 12:13 AM.
HDTVChallenged is offline  
post #42 of 55 Old 09-16-2014, 06:15 AM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
Quote:
Originally Posted by HDTVChallenged View Post
You may be waiting for a long time. To quote Mr. Spock, 'To have a meaningful discussion of the subject, one must first have a common frame of reference.'

To point you in the right direction, you might want to try plotting the REC2020 color space (the definitions of primary coordinates in CIE31) along with the corresponding primary coordinates of the REC709 and REC601 color spaces. From this exercise, you *should* find that REC2020 can easily produce *all* of the colors present in REC709 and REC601. Therefore, it is a trivial mathematical problem to map REC709 or REC601 colors into REC2020 space.

It's pointless to "debate" this further ... either you understand the issue, or you need to do a little more study ... and reasoning.

PS: It's just math (linear algebra specifically and a little geometry to help out with the visualization.)
Your assumptions are that the engineers do their jobs properly and the equipment works flawlessly. It is said, "Things that work on the drawing board do not always work in practice.

The debate has mutated far from the original intent of the thread. I do understand the math.

Due to your lack of interest in forwarding any information of REC2020 being a "all in one color space", one can only assume that the information you had stated prior is bogus. This conclusion is only amplified by your arrogant response. Your faith in your mathematical hypotheses is admirable but the concept of this forum is information exchange. If you were conveying the truth than forwarding the information would be no problem. Unless you are willing to support your position, don't state it.

Last edited by randal_r; 09-16-2014 at 08:20 AM.
randal_r is online now  
post #43 of 55 Old 09-16-2014, 08:04 AM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
Quote:
Originally Posted by spacediver View Post
My comments were restricted purely to gamut considerations, not things like bit depth, resolution, frame rate, or luma coefficients. Let me try to be clearer. If you have an RGB test pattern that instructs the display to only "light up" one channel, then the color that the display renders will depend entirely on the display's native primaries (assuming no color management is in place). If you have an sRGB display and a display that somehow had the rec 2020 primaries, then the resulting image will reflect the gamut of the display.



This is your original quote:

You make it seem as if rec2020 displays will somehow make older material obsolete as one would need to convert (remaster) the material. So long as a rec 2020 display can remap its color gamut to Rec 709, then there will be no difference viewing content on that remapped display as there is on current Rec 709 displays (assuming of course that all other Rec 709 specs are able to be emulated on this future display).



Again, I was only talking about gamut considerations, not resolution or color depth.
I am not saying that older material would not work on a REC2020 display; I am saying that the quality of material that is not of a REC709 and of a lower resolution would be so degraded that to make the material render a unreasonable image, where remastering would probably be the only option. The 4k displays render a good image when playing 4K and 2K material (upscaled or not) but try putting on a SD material (Television Signal), and compare the quality. Now lets project further, the degradation is amplified when utilizing an 8K display. To make the material appealing, remastering might be the only way. This has already been the case in the present day with movies such as "Casablanca". This movie has been remastered to an HD format and may play on a REC2020 display as others have suggested. However, there is much more material of lower resolution that has not been remastered and will need to be in order to be playable on such displays.

This industry, as well as others, always have legacy issues. Would it not be easier to have all cameras and displays utilize a digital form of gamma rather then an analog version and resolve the gamma issue once and for all? The reason it will never happen are many, such as due to older equipment throughout the world (cameras for example); in a word, "Legacy".

A question, can a display of today with it's color gamut fixed at REC709 display REC2020 material? Can media storage devices of today be able to play REC2020 material? Even the the "wide" gamuts do not come anywhere near the REC2020 gamut. I have not ever heard of anyone expanding the native gamut beyond it's native settings. Some out there have put blind faith in the remapping abilities of a display that is at best still only a prototype. I would put out this question; has anyone actually verified the remapping abilities of a REC2020 display? One individual stressed on my ignorance of the subject. His faith in the mathematics of the subject is firm and because I did not share his view of the mathematical hypotheses, I was deemed stupid and ignorant by him. I have been around long enough to know that theory and practice can be two different things. All through my university days we were taught to question everything, to make no assumptions. The technology of the remapping appears to work but when business makes economical decisions, technology suffers and is compromised for the all mighty dollar. Many displays of today are plagued with technological issues due to economical decisions. There are others out there that are just poorly designed. I feel that with a product with such a low profit margin and the insatiable level of greed that exists in business, I do not think this issue will ever disappear.
randal_r is online now  
post #44 of 55 Old 09-16-2014, 10:34 AM
AVS Forum Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,936
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 215
Quote:
Originally Posted by randal_r View Post
Your assumptions are that the engineers do their jobs properly and the equipment works flawlessly. It is said, "Things that work on the drawing board do not always work in practice..
All I said was that an REC2020 display would have no problems displaying REC709 or REC601 as originally intended. The presumption is that you actually have a working REC2020 display ... or at least one that comes very close.
HDTVChallenged is offline  
post #45 of 55 Old 09-16-2014, 10:42 AM
AVS Forum Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,936
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 215
PS: Yes there's no question that 480i/SD material will look "horrible" (resolution-wise, et al.) on an REC2020/4k/8k display. Well Duh! .... What do you expect? That's a different question than whether or not REC709/601 colors can be rendered properly on an REC2020 display.

All I and others are saying is that there's no "need" to remaster REC709/601(SMPTE-C actually) material for display on REC2020. We're *not* saying that a REC2020 display can magically turn 480i/SD into 4k/8k.

Last edited by HDTVChallenged; 09-16-2014 at 10:46 AM.
HDTVChallenged is offline  
post #46 of 55 Old 09-16-2014, 11:00 AM
AVS Forum Special Member
 
HDTVChallenged's Avatar
 
Join Date: Jan 2002
Posts: 8,936
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 215
Quote:
Originally Posted by randal_r View Post

A question, can a display of today with it's color gamut fixed at REC709 display REC2020 material? Can media storage devices of today be able to play REC2020 material? Even the the "wide" gamuts do not come anywhere near the REC2020 gamut. I have not ever heard of anyone expanding the native gamut beyond it's native settings. Some out there have put blind faith in the remapping abilities of a display that is at best still only a prototype. I would put out this question; has anyone actually verified the remapping abilities of a REC2020 display? One individual stressed on my ignorance of the subject. His faith in the mathematics of the subject is firm
What part of statement, "REC709 and REC601/SMPTE-C colorspaces are fully contained sub-sets of the REC2020 color-space," do you not understand?
HDTVChallenged is offline  
post #47 of 55 Old 09-16-2014, 11:03 AM
AVS Forum Special Member
 
zoyd's Avatar
 
Join Date: Sep 2006
Location: Planet Dog
Posts: 6,293
Mentioned: 88 Post(s)
Tagged: 0 Thread(s)
Quoted: 1315 Post(s)
Liked: 1051
I've got a display sitting right here with native primaries just outside Adobe RGB which has no problem emulating either that space or Rec709.
zoyd is offline  
post #48 of 55 Old 09-16-2014, 02:53 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,034
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 85
Quote:
Originally Posted by randal_r View Post
I am not saying that older material would not work on a REC2020 display; I am saying that the quality of material that is not of a REC709 and of a lower resolution would be so degraded that to make the material render a unreasonable image, where remastering would probably be the only option. The 4k displays render a good image when playing 4K and 2K material (upscaled or not) but try putting on a SD material (Television Signal), and compare the quality.
Are you suggesting that SD material upscaled on a 2k/4k/8k display will look worse than SD material on an "SD" display?

Quote:
Originally Posted by randal_r View Post
This industry, as well as others, always have legacy issues. Would it not be easier to have all cameras and displays utilize a digital form of gamma rather then an analog version and resolve the gamma issue once and for all? The reason it will never happen are many, such as due to older equipment throughout the world (cameras for example); in a word, "Legacy".
What do you mean by digital vs. analog gamma? Are you referring to the CRT's "native" gamma when you say analog?

Quote:
Originally Posted by randal_r View Post
A question, can a display of today with it's color gamut fixed at REC709 display REC2020 material? Can media storage devices of today be able to play REC2020 material? Even the the "wide" gamuts do not come anywhere near the REC2020 gamut. I have not ever heard of anyone expanding the native gamut beyond it's native settings.
That's because it's physically impossible to expand a gamut beyond it's native gamut. You can't expand a Rec709 gamut to anything larger than Rec709. But to use this logic to imply that you can't "shrink" a larger gamut to render a smaller one is misguided.
spacediver is offline  
post #49 of 55 Old 09-16-2014, 05:35 PM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
Quote:
Originally Posted by spacediver View Post
Are you suggesting that SD material upscaled on a 2k/4k/8k display will look worse than SD material on an "SD" display?
Yes


What do you mean by digital vs. analog gamma? Are you referring to the CRT's "native" gamma when you say analog?

Yes, verses a straight line from 0 to 1 for luminance


That's because it's physically impossible to expand a gamut beyond it's native gamut. You can't expand a Rec709 gamut to anything larger than Rec709. But to use this logic to imply that you can't "shrink" a larger gamut to render a smaller one is misguided.
The closer you move the primaries closer to the white point the more the image will take on the properties of B&W; Color will diminish. This is how the color control function works.

Last edited by randal_r; 09-16-2014 at 05:41 PM.
randal_r is online now  
post #50 of 55 Old 09-16-2014, 05:57 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,034
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 85
Quote:
Originally Posted by randal_r View Post
Yes
I'm not an expert in scaling, but as I understand it, with 4k or 8k there should be a variety of scaling solutions. If, for example, you map each pixel of the SD content onto four pixels of the higher res display, the image should look pretty much identical. What makes you think it will look worse?

Quote:
Originally Posted by randal_r View Post
Yes, verses a straight line from 0 to 1 for luminance
With 8 bits, linear light would be disastrous. I suggest you do some reading. Here and here.

Quote:
Originally Posted by randal_r View Post
The closer you move the primaries closer to the white point the more the image will take on the properties of B&W; Color will diminish. This is how the color control function works.
But that's exactly what you want when you're remapping a wide gamut to the Rec 709 or Rec 601 primaries.
spacediver is offline  
post #51 of 55 Old 09-16-2014, 07:18 PM
Senior Member
 
randal_r's Avatar
 
Join Date: Mar 2010
Location: Ontario Canada
Posts: 411
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 168 Post(s)
Liked: 113
Quote:
Originally Posted by spacediver View Post
I'm not an expert in scaling, but as I understand it, with 4k or 8k there should be a variety of scaling solutions. If, for example, you map each pixel of the SD content onto four pixels of the higher res display, the image should look pretty much identical. What makes you think it will look worse?



With 8 bits, linear light would be disastrous. I suggest you do some reading. Here and here.



But that's exactly what you want when you're remapping a wide gamut to the Rec 709 or Rec 601 primaries.
Let me explain how I see the world. With regards to mapping you said you would map each pixel of the SD content onto four pixels of the higher res display. With the primary SD pixel you would have to add three or more filler pixels. the addition of filler pixels contribute to a softer image. The higher resolution you are trying to achieve, the greater the degradation. Have you ever seen video material form the 50's? Why does it look so bad. Even with the television station enhancing it as much as they can the image looks degraded. On some HD channels showing SD material the image is presented as a smaller image with a black boarder. Some television stations will take that smaller image and zoom on the image to simulate an HD image as much as possible.


Now to gamma. I have attached a file to illustrate digital vs analog. Note the left chart, the straight line going from the bottom left to the upper right represents digital. This has a 1:1 rise. Note the other curved lines, these represent analog. Not wanting to repeat what was taught in the THX class I will stop here.


Remapping from between bt709, bt601, ebu and smpt-c is not much of an issue. Based on the attached gamut chart you can see that the aforementioned color spaces are almost identical in size merely the primaries have different coordinates. When it comes to displays that have larger gamuts unless the display has internal circuitry to handle REC2020 (for example), and REC709 Issues with color will occur. Please note that when I refer to displays I am not referring to computer monitors. As I have stated before, when primaries are moved closer to the white point, the more color diminishes. Much of what has been expressed to me regarding remapping is suggested that the display device will handle this on the fly. Remapping is not like upscaling. Remapping as it has been suggested to me of a gamut would require determining the parameters of the material being imputed. Adjusting the resolution parameters (to upscale or not) and finally to adjust the color space such as primary and seconday color point cooridinates, adjusting issues of hue, saturation and color. I personally have not heard or seen a display that does this (if there is please bring it to my attention), to date. This is why I suggest remastering of material might be in order.
Attached Thumbnails
Click image for larger version

Name:	GammaCharts.png
Views:	210
Size:	180.4 KB
ID:	265841   Click image for larger version

Name:	Screenshot-6.png
Views:	1170
Size:	266.8 KB
ID:	265921  

Last edited by randal_r; 09-16-2014 at 07:23 PM.
randal_r is online now  
post #52 of 55 Old 09-16-2014, 07:57 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,034
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 85
Quote:
Originally Posted by randal_r View Post
Let me explain how I see the world. With regards to mapping you said you would map each pixel of the SD content onto four pixels of the higher res display. With the primary SD pixel you would have to add three or more filler pixels.
No there is no interpolation or filling in, as you put it. You simply replicate the pixels. In other words, suppose your original content was mastered for a 2x2 pixel display, and the image was a checkerboard pattern, with two white pixels and two black pixels. If you want to render this image on a 4x4 pixel display, you would simply replicate each of the original pixels four times. So now you end up with exactly the same image, but each checker patch is now represented by four pixels. No filling in or interpolation necessary. We are not "adding" detail, we are simply replicating the detail faithfully.

Quote:
Originally Posted by randal_r View Post
Now to gamma. I have attached a file to illustrate digital vs analog. Note the left chart, the straight line going from the bottom left to the upper right represents digital. This has a 1:1 rise. Note the other curved lines, these represent analog. Not wanting to repeat what was taught in the THX class I will stop here.
I understood exactly what you meant, and I still haven't a clue why you refer to the linear gamma as "digital". I have a feeling you're conflating multiple issues here. Regardless of whether you call it digital or not, you don't seem to understand why a 1:1 rise would be disastrous with anything less than 16 bits. I forwarded you some important information that explains this quite well.

To put it in a nutshell, the human visual response to luminance is very nonlinear. Generally we are more sensitive to luminance changes in low luminance region, and less sensitive at higher luminances. If we used a "1:1 gamma", and only had 8 bits (say 256 steps to encode the dynamic range), the first few video levels would exhibit noticeable banding. With such few bits to work with, you need a nonlinear luminance function to make perceptually efficient use of the bits, and to maintain perceptual uniformity across the dynamic range. It just so happens that the triode transfer functions of factory calibrated BVM CRTs happened to have a gamma that was very nearly the inverse of the human lightness response, which made it an ideal gamma to work with.

I know first hand how a linear gamma affects image quality, as for my psychophysical research I often calibrate the displays for a linear gamma (1:1), so that contrast manipulations which are done on the pixel value domain are naturally scaled relative to luminance.

From the first link I provided in my previous post (note the bolded sections):

Quote:
Originally Posted by Charles Poynton
The first section of this note described the nonlinear response of human vision that relates luminance to perceived lightness. The second section described how the nonlinear transfer function of a CRT relates a voltage signal to intensity. Here's the amazing coincidence: the CRT voltage-to-intensity function is very nearly the inverse of the luminance-to-lightness relationship of vision. That means that coding a luminance signal into voltage, to be turned back into luminance by a CRT, is very nearly the optimal coding to minimize the perceptibility of noise that is introduced into the signal. CRT voltage is remarkably perceptually-uniform.

Suppose you have a luminance value that is determined quite precisely, but you must use a channel having only eight bits to convey that value to a distant observer. Consider a linear light representation with eight bits, where code zero represents black and code 255 represents white. Code value 100 represents a shade of gray that is approximately at the perceptual threshold: for codes above 100, the ratio between intensity values of adjacent codes is less than one percent, and for codes below 100, the ratio between intensity values of adjacent codes is greater than one percent.

Luminance codes below 100 suffer increasing artifacts as the code value decreases towards black, due to the visibility of the luminance difference between adjacent codes: at code 50, the ratio between adjacent codes is 2%, which is noticeable to most observers. These artifacts are especially objectionable in pictures having large areas of smoothly-varying shades.

Luminance codes above 100 suffer no artifacts due to the visibility of the jumps between codes. However, as the code value increases towards white, the codes have decreasing perceptual utility. For example, at code 200, the ratio between adjacent intensity steps is half a percent, well below the threshold of visibility. Codes 200 and 201 are visually indistinguishable: code 201 is perceptually useless, and could be discarded without being noticed. This example shows that a linear-luminance representation is a poor choice for an eight-bit channel
Quote:
Originally Posted by randal_r View Post
Remapping from between bt709, bt601, ebu and smpt-c is not much of an issue. Based on the attached gamut chart you can see that the aforementioned color spaces are almost identical in size merely the primaries have different coordinates. When it comes to displays that have larger gamuts unless the display has internal circuitry to handle REC2020 (for example), and REC709 Issues with color will occur.
This is why in my original post i said:

Quote:
Originally Posted by spacediver View Post
A good wide gamut display would have the ability to remap the primaries depending on source material, either through metadata or through the user's manual control
Any wide gamut display worth its salt can fluidly remap its primaries. For example, nowadays we are seeing studio monitors that effortlessly remap between DIC P3 and Rec 709.

Quote:
Originally Posted by randal_r View Post
As I have stated before, when primaries are moved closer to the white point, the more color diminishes.
I'm not sure why you bring this up. You do realize that Rec 709 is, by definition, less colorful than Rec 2020. So obviously if you are remapping the primaries from Rec 2020 to Rec 709, the primaries are going to be less saturated. But this is exactly what you want if you're watching content that is mastered in Rec 709!

Last edited by spacediver; 09-17-2014 at 07:23 PM.
spacediver is offline  
post #53 of 55 Old 09-16-2014, 08:17 PM
AVS Forum Special Member
 
zoyd's Avatar
 
Join Date: Sep 2006
Location: Planet Dog
Posts: 6,293
Mentioned: 88 Post(s)
Tagged: 0 Thread(s)
Quoted: 1315 Post(s)
Liked: 1051
Quote:
Originally Posted by spacediver View Post
No there is no interpolation or filling in, as you put it. You simply replicate the pixels. In other words, suppose your original content was mastered for a 2x2 pixel display, and the image was a checkerboard pattern, with two white pixels and two black pixels. If you want to render this image on a 4x4 pixel display, you would simply replicate each of the original pixels four times. So now you end up with exactly the same image, but each checker patch is now represented by four pixels. No filling in or interpolation necessary. We are not "adding" detail, we are simply replicating the detail faithfully.
Nicely put, another way to describe this is called oversampling. For a given screen size there will be absolutely no visual difference between SD material rendered at 2K to that same source rendered at 4K or 8K.

Quote:
Any wide gamut display worth its salt can fluidly remap its primaries. For example, nowadays we are seeing studio monitors that effortlessly remap between DIC P3 and Rec 709.
Yes, the processing to do this is trivial and present even in consumer displays as 6 point LUTs. What might be more difficult is linearizing the display response across the wider gamut but that's not the discussion here, downsizing to rec709 will not be an issue.
zoyd is offline  
post #54 of 55 Old 09-16-2014, 09:54 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,034
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 85
Quote:
Originally Posted by randal_r View Post
Now to gamma. I have attached a file to illustrate digital vs analog. Note the left chart, the straight line going from the bottom left to the upper right represents digital. This has a 1:1 rise. Note the other curved lines, these represent analog.
I just looked more closely at that image - the one on the left seems to depict different LUT schemes. The linearized LUT simply means that the video input (0-255) is converted into a signal (whether that signal be digital or analog) that is linear. The relationship between LUT input and output (left image) is rather different between the relationship between input video level and luminance. Perhaps this is part of the confusion?
spacediver is offline  
post #55 of 55 Old 09-24-2014, 12:13 PM - Thread Starter
AVS Forum Special Member
 
nuke's Avatar
 
Join Date: Jul 2001
Location: CA, USA
Posts: 1,290
Mentioned: 57 Post(s)
Tagged: 0 Thread(s)
Quoted: 242 Post(s)
Liked: 138
Thanks Ron,

I was goofing around with the files you generated and it turns out that Sony FMPX10 4k player is very picky about what it will play. It will only play 8-bit files and it will only play files with .mp4 extensions. It pretty much refused to play them. I did a little hacking trying to convert them both to something it would recognize or to something it would play, but didn't really get far with them.




Quote:
Originally Posted by dr1394 View Post
I've updated the pattern with slightly more accurate YCbCr values (by using all 10-bits instead of just shifting 8-bit values). Also, since it compresses to a fairly small file, I've extended the play length to 5 minutes.

http://www.w6rz.net/movingbaruhd5.mkv

Here's the raw H.264 bitstream in case it needs to be muxed into another container.

http://www.w6rz.net/movingbaruhd5.264

You may have to right click on the links and select "Save Link As..." to properly download the file.

Here's the source code (that shows the YCbCr values used).

http://www.w6rz.net/movingbaruhd.c

The shift left by 6 is to satisfy the "yuv422p16le" input format used by x264. I'm not sure where that comes from, since the H.264 reference decoder doesn't do that. The x264 command line is:
Code:
x264 -o movingbaruhd.mkv movingbaruhd.yuv --input-res 2880x2160 --input-csp yuv422p16le --demuxer lavf --fps 24 --sar 4:3 --colorprim bt2020 --transfer bt2020-10 --colormatrix bt2020nc
Here's the BT.2020 signalling in the H.264 bitstream.
Code:
   vui_parameters_present_flag: 1
    aspect_ratio_info_present_flag: 1
     aspect_ratio_idc:14
    overscan_info_present_flag: 0
    video_signal_info_present_flag: 1
     video_format: 5
     video_full_range_flag: 0
     colour_description_present_flag: 1
      colour_primaries: 9
      transfer_characteristics: 14
      matrix_coefficients: 9
    chroma_loc_info_present_flag: 0
colour primaries = 9 signals BT.2020 primaries
transfer_characteristics = 14 signals BT.2020 gamma for a 10 bit system
matrix_coefficients = 9 signals BT.2020 non-constant luminance system

For creating test patterns, the only parameter that matters is the matrix coefficients (since these are used to convert RGB to YCbCr). Primaries and gamma are entirely up to the display to handle properly.

Ron

nuke
nuke is offline  
Sponsored Links
Advertisement
 
Reply Display Calibration
Gear in this thread



Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off