Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Announcements and Chat > General Discussion

Reply
 
Thread Tools Search this Thread Display Modes
Old 8th March 2013, 17:24   #1  |  Link
jq963152
Registered User
 
Join Date: Apr 2012
Posts: 691
4:2:0 vs. 4:2:2 vs. 4:4:4

Quote:
Originally Posted by pieter3d View Post
The human eye is not nearly as sensitive to color as it is to texture. So I wouldn't view the omission of the higher fidelity chroma profiles in this first version as disappointing.
IMHO, the following comparison for example:


Quote:

4:4:4:



4:2:0:


is very much disappointing .



Chroma subsampling (as well as interlacing) are very evil things IMHO. They both should have been abandoned a long time ago IMHO .

Last edited by jq963152; 28th December 2016 at 20:40.
jq963152 is offline   Reply With Quote
Old 8th March 2013, 18:01   #2  |  Link
pieter3d
Registered User
 
Join Date: Jan 2013
Location: Santa Clara CA
Posts: 116
That is quite the cherry-picked example. For natural content (stuff from cameras) it is nearly imperceptible, especially for HD stuff. And it gets you 50% data rate reduction before you even begin compressing, so don't count on it disappearing any time soon. See here: http://users.wfu.edu/matthews/misc/jpg_vs_gif/JpgCompTest/JpgChromaSub.html
pieter3d is offline   Reply With Quote
Old 8th March 2013, 18:56   #3  |  Link
jq963152
Registered User
 
Join Date: Apr 2012
Posts: 691
Quote:
Originally Posted by pieter3d View Post
And it gets you 50% data rate reduction before you even begin compressing

But x264 doesn't need twice the bitrate when going from 4:2:0 to 4:4:4?

Quote:
Originally Posted by pieter3d View Post
For natural content (stuff from cameras) it is nearly imperceptible, especially for HD stuff. See here: http://users.wfu.edu/matthews/misc/jpg_vs_gif/JpgCompTest/JpgChromaSub.html

It's definitely not "nearly imperceptible" IMHO .

Quote:
Originally Posted by pieter3d View Post
so don't count on it disappearing any time soon.


Last edited by jq963152; 8th March 2013 at 19:00.
jq963152 is offline   Reply With Quote
Old 8th March 2013, 19:01   #4  |  Link
pieter3d
Registered User
 
Join Date: Jan 2013
Location: Santa Clara CA
Posts: 116
I mean twice the bitrate of the raw content. You can compress the chroma heavier in 4:4:4, so it doesn't end up being a full 2x bitrate anyway, especially since in natural content, chroma samples are pretty smooth textures and easy to predict. This is also why you might as well subsample. Try running x264 on content like your red/blue text example.
pieter3d is offline   Reply With Quote
Old 8th March 2013, 19:10   #5  |  Link
jq963152
Registered User
 
Join Date: Apr 2012
Posts: 691
Quote:
Originally Posted by pieter3d View Post
I mean twice the bitrate of the raw content.
This thread is about H.265. So, it's not about the raw content .

And who cares about the size of the raw content ?

4:4:4 looks so much better, so if the raw content is a little bit bigger, so be it .

IMHO
jq963152 is offline   Reply With Quote
Old 8th March 2013, 19:12   #6  |  Link
pieter3d
Registered User
 
Join Date: Jan 2013
Location: Santa Clara CA
Posts: 116
Quote:
Originally Posted by jq963152 View Post
4:4:4 looks so much better, so if the raw content is a little bit bigger, so be it .

IMHO
I bet you'd be hard-pressed to tell the difference between a 4k movie at regular viewing distance in 420 vs 444.
pieter3d is offline   Reply With Quote
Old 8th March 2013, 19:55   #7  |  Link
jq963152
Registered User
 
Join Date: Apr 2012
Posts: 691
Quote:
Originally Posted by pieter3d View Post
I bet you'd be hard-pressed to tell the difference between a 4k movie at regular viewing distance in 420 vs 444.
4:2:0 vs. 4:4:4 is definitly very noticeable on videogame footage for example IMHO.

Also, regardless if noticeable or not: hey, we're in the year 2013, FULL-HD should be true FULL-HD and not just 1080p luma and a mere 540p chroma...

And so on...

Also, by the way:

Quote:
Originally Posted by pieter3d View Post
See here: http://users.wfu.edu/matthews/misc/jpg_vs_gif/JpgCompTest/JpgChromaSub.html

The 4:4:4 example over there looks much better than the 4:2:0 counterpart IMHO. Much better. It looks much cleaner and therefore has much better definition.

And look at that mess in the magnified picture .

As already mentioned, chroma subsampling is a very evil thing .
jq963152 is offline   Reply With Quote
Old 8th March 2013, 21:04   #8  |  Link
Sagittaire
Testeur de codecs
 
Sagittaire's Avatar
 
Join Date: May 2003
Location: France
Posts: 2,420
Quote:
Originally Posted by jq963152 View Post
4:2:0 vs. 4:4:4 is definitly very noticeable on videogame footage for example IMHO.

Also, regardless if noticeable or not: hey, we're in the year 2013, FULL-HD should be true FULL-HD and not just 1080p luma and a mere 540p chroma...

And so on...

Also, by the way:




The 4:4:4 example over there looks much better than the 4:2:0 counterpart IMHO. Much better. It looks much cleaner and therefore has much better definition.

And look at that mess in the magnified picture .

As already mentioned, chroma subsampling is a very evil thing .

yes it's true 4:4:4 is really better visualy than 4:2:0. Anyway you must make comparison at same bytes size. And in this case for Human Visual System (HVS) 4:2:0 will be really better than 4:4:4. For HVS luma information is really more important than chroma information. By the way use 4:2:0 system is already by definition a first and simple HVS picture compression format.

For illustrate this information:

- with RGB24 format (8 bits for each chroma chanel) you have 720x576 pixels and 24 bits for each pixel: 720x576x24 = 10 Mbits for uncompressed RGB24 picture.

- with YV12 format you have 12 bits for each pixel. Wiith this format you can have in practice picture at 1280x720x12 = 11MBits for uncompressed YV12 picture.

And it's really easy for me to say that at same uncompressed size 720p YV12 picture will be visually always really better than 576p RGB24 picture.
__________________
Le Sagittaire ... ;-)

1- Ateme AVC or x264
2- VP7 or RV10 only for anime
3- XviD, DivX or WMV9

Last edited by Sagittaire; 8th March 2013 at 21:27.
Sagittaire is offline   Reply With Quote
Old 8th March 2013, 21:30   #9  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 666
Game captures are just one type of source, and not nearly a common case (I would say that remote desktop functionality is much more important than them). As pieter3d said, the difference from chroma subsampling is hardly visible for natural content, like... movies and tv series, online video, camcorder, webcam and phone footage (95 % of stuff that people distribute), especially at 1080p and 4K.
mandarinka is offline   Reply With Quote
Old 8th March 2013, 22:12   #10  |  Link
jq963152
Registered User
 
Join Date: Apr 2012
Posts: 691
Quote:
Originally Posted by Sagittaire View Post
And it's really easy for me to say that at same uncompressed size 720p YV12 picture will be visually always really better than 576p RGB24 picture.
You might have a point there.

But you forgot to mention, that, at 1280x720 4:2:0, chroma is only at 640x360, which IMHO is ridiculous for high definition in the year 2013 .

Come on, 640x360 (1280x720 4:2:0), IMHO, is not high definition . Nor is 960x540 (1920x1080 4:2:0).

IMHO it's also a fail to move from 1080p to 4K but essentially leaving chroma at just 1920x1080 (3840x2160 4:2:0).

Quote:
Originally Posted by mandarinka View Post
As pieter3d said, the difference from chroma subsampling is hardly visible for natural content, like... movies and tv series, online video, camcorder, webcam and phone footage (95 % of stuff that people distribute), especially at 1080p and 4K.
Chroma subsampling makes the picture blurry/muddy. IMHO this can't be described as "hardly visible".

Just look at the picture examples in the link from "pieter3d" for example. 4:4:4 looks a lot cleaner .

IMHO, "High Definition" should be full high definition and not just luma high definition + only chroma low definition .

Last edited by jq963152; 8th March 2013 at 22:15.
jq963152 is offline   Reply With Quote
Old 9th March 2013, 00:39   #11  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 666
Quote:
Originally Posted by jq963152 View Post
Chroma subsampling makes the picture blurry/muddy.
You keep saying that, but it only happens in your game footage, particularly low resolution one like from old arcades. Go try to test with some high resolution footage, like Parkjoy... it has enough of excess resolution for testing for example 1080P 4:4:4 and 4:2:0 IIRC.

Numerical argumenting that "but the chroma resolution is half that!" only makes so much sense...
mandarinka is offline   Reply With Quote
Old 9th March 2013, 04:06   #12  |  Link
STaRGaZeR
4:2:0 hater
 
Join Date: Apr 2008
Posts: 1,307
Quote:
Originally Posted by mandarinka View Post
As pieter3d said, the difference from chroma subsampling is hardly visible for natural content, like... movies and tv series, online video, camcorder, webcam and phone footage (95 % of stuff that people distribute), especially at 1080p and 4K.
Anything red/blue in any of the content you cited disagrees with that, being letters, logos, clothes, etc.

You can get away with it because in most content there is little chroma information to begin with, but as soon as the content has chroma detail 4:2:0 goes to hell: resolution is four times lower compared to luma.

Chroma subsampling is an aberration from the analog era used to save bandwidth and hence costs. Since in analog you actually work with "uncompressed" signals all the time, it makes perfect sense to butcher the relatively unimportant chroma information and cut the needed bandwidth in half while maintaining very good quality overall. This is absolutely not the case with compressed digital video.

It's the same situation we have with interlacing, and in an ideal world they should be given equal treatment: the middle finger. The video industry is far from ideal unfortunately.
__________________
Specs, GTX970 - PLS 1440p@96Hz
Quote:
Originally Posted by Manao View Post
That way, you have xxxx[p|i]yyy, where xxxx is the vertical resolution, yyy is the temporal resolution, and 'i' says the image has been irremediably destroyed.
STaRGaZeR is offline   Reply With Quote
Old 9th March 2013, 09:12   #13  |  Link
Poutnik
Useless idea generator
 
Poutnik's Avatar
 
Join Date: Apr 2004
Location: Europe, Czech Republic, Brno
Posts: 354
Quote:
Originally Posted by STaRGaZeR View Post

Chroma subsampling is an aberration from the analog era used to save bandwidth and hence costs.
1:1:1 sampling is unbalanced, wrt to human resolution, as the eye is using colour subsampling as well.

Bigger sense than removing subsampling
has improving overall resolution to keep pixel viewing angles under angle eye resolution.
__________________
Vista64 Premium SP2 / C2D E4700 2.6GHz/ 6GB RAM/ Intel GMA 3100 / DTV Leadtek DONGLE GOLD USB2 /
focused to DVB-T MPEG2 PS capture -> ProjectX -> M2V/MP2 -> MeGUI/AVS -> MP4[AVC/AAC]

Last edited by Poutnik; 9th March 2013 at 09:21.
Poutnik is offline   Reply With Quote
Old 9th March 2013, 20:37   #14  |  Link
STaRGaZeR
4:2:0 hater
 
Join Date: Apr 2008
Posts: 1,307
Quote:
Originally Posted by Poutnik View Post
1:1:1 sampling is unbalanced, wrt to human resolution, as the eye is using colour subsampling as well.
Human vision doesn't work in pixels, and what the eye sees is not what your brain interprets afterwards. Plus you're ignoring display technology and the need to upsample to 4:4:4 in the middle. In real life colors next to a sharp edge (detail) don't bleed into each other, they do with choma subsampling because of lack of resolution.

In the digital world blindly removing samples is the worst possible way of reducing bandwidth if what you seek is quality.

Quote:
Originally Posted by Poutnik View Post
Bigger sense than removing subsampling
has improving overall resolution to keep pixel viewing angles under angle eye resolution.
The two are in no way mutually exclusive.
__________________
Specs, GTX970 - PLS 1440p@96Hz
Quote:
Originally Posted by Manao View Post
That way, you have xxxx[p|i]yyy, where xxxx is the vertical resolution, yyy is the temporal resolution, and 'i' says the image has been irremediably destroyed.
STaRGaZeR is offline   Reply With Quote
Old 9th March 2013, 23:50   #15  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 666
If 4:4:4 encode ends up having a higher bitrate, they are. You could use savings from 4:2:0 to increase overal resolution, and people are more sensitive to luma.
BTW, this is fairly off-topic already.
mandarinka is offline   Reply With Quote
Old 10th March 2013, 00:52   #16  |  Link
Poutnik
Useless idea generator
 
Poutnik's Avatar
 
Join Date: Apr 2004
Location: Europe, Czech Republic, Brno
Posts: 354
Quote:
Originally Posted by STaRGaZeR View Post
Human vision doesn't work in pixels, and what the eye sees is not what your brain interprets afterwards.
What else are cones, if not pixels, even if not arranged in rectangular net ? And eye provide to brain higher resolution for intensity than for color.

Quote:
Plus you're ignoring display technology and the need to upsample to 4:4:4 in the middle. In real life colors next to a sharp edge (detail) don't bleed into each other, they do with choma subsampling because of lack of resolution.
If there is sufficient resolution you do not notice bleeding.
Quote:

In the digital world blindly removing samples is the worst possible way of reducing bandwidth if what you seek is quality.
Far from blindly.


The two are in no way mutually exclusive.[/QUOTE]
__________________
Vista64 Premium SP2 / C2D E4700 2.6GHz/ 6GB RAM/ Intel GMA 3100 / DTV Leadtek DONGLE GOLD USB2 /
focused to DVB-T MPEG2 PS capture -> ProjectX -> M2V/MP2 -> MeGUI/AVS -> MP4[AVC/AAC]
Poutnik is offline   Reply With Quote
Old 12th March 2013, 10:46   #17  |  Link
ajp_anton
Registered User
 
ajp_anton's Avatar
 
Join Date: Aug 2006
Location: Stockholm/Helsinki
Posts: 758
But if the content doesn't have much chroma detail, the encoder should be able to see this and compress it heavier to the point where it takes the same space as subsampled chroma would have. And if the content has chroma detail (computer/game graphics), the bitrate might go up but so does the quality in a way that's simply not possible with 4:2:0.
Also, some areas of the same image might have high frequency detail while others don't. Better to let the encoder decide where to spend its bits, rather than recuce the quality uniformly. There's a reason we have advanced algos like x264's CRF and AQ, nobody uses QP-mode.

So I guess the only real problem is that it's difficult to compress an upscaled image as well as the original, and the assumption is that chroma "in the real world" is low on detail like an upscale.

Last edited by ajp_anton; 12th March 2013 at 10:49.
ajp_anton is offline   Reply With Quote
Old 13th March 2013, 22:49   #18  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 2,442
Quote:
Originally Posted by ajp_anton View Post
But if the content doesn't have much chroma detail, the encoder should be able to see this and compress it heavier to the point where it takes the same space as subsampled chroma would have. And if the content has chroma detail (computer/game graphics), the bitrate might go up but so does the quality in a way that's simply not possible with 4:2:0.
Correct to a point. But just like there's a point in luma where bits/pixel drops low enough that a lower frame size looks better, this could happen in the chroma channel as well.

Another reason why 4:2:0 was popular is that it cut memory bandwidth requirements in half (12 bits per pixel instead of 24), which was a big deal for a long time.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Instant Video

My Compression Book

Amazon Instant Video is hiring! PM me if you're interested.
benwaggoner is offline   Reply With Quote
Old 14th March 2013, 01:23   #19  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,314
It's still important; 4K uncompressed video at 24fps at full chroma resolution is 1.1 GB/s, and at 60fps is 2.9. That's a pretty heavy load on the underlying bus and transfer links, especially if any processing is being done on it, and having to transfer only half of that can be a big deal.

But that's no reason why it shouldn't be specified and supported, since current hardware can easily cope with it at lower resolutions and lower resolutions are where it matters more. HEVC is designed entirely around the needs of HD, not bringing forward SD, which shouldn't be too surprising; color depth is far more important for HD than chroma sampling, so Main 10 is in the initial spec while range extensions won't be until some unknown future date.
__________________
There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order. ~ Ed Howdershelt
foxyshadis is offline   Reply With Quote
Old 14th March 2013, 10:34   #20  |  Link
jq963152
Registered User
 
Join Date: Apr 2012
Posts: 691
Quote:
Originally Posted by Poutnik View Post
as the eye is using colour subsampling as well.
Quote:
Originally Posted by Poutnik View Post
What else are cones, if not pixels, even if not arranged in rectangular net ? And eye provide to brain higher resolution for intensity than for color.
All invalid points to me, because i can perfectly see the difference between 4:2:0 and 4:4:4 with, guess what, my own eyes .

And it's a HUGE difference i tell ya .

Quote:
Originally Posted by Poutnik View Post
If there is sufficient resolution you do not notice bleeding.
I always notice the bleeding. And it's very annoying.

Quote:
Originally Posted by foxyshadis View Post
But that's no reason why it shouldn't be specified and supported
IMHO, 4:4:4 should be mandatory and the be the only supported format. 4:2:0 (or 4:2:2) shouldn't be supported any longer IMHO.

It's quite sad to see how many of you in here seem to be okay with chroma subsampling...

Last edited by jq963152; 14th March 2013 at 10:36.
jq963152 is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

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

Forum Jump


All times are GMT +1. The time now is 04:40.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2018, vBulletin Solutions Inc.