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 > Video Encoding > High Efficiency Video Coding (HEVC)

Reply
 
Thread Tools Search this Thread Display Modes
Old 4th December 2015, 18:37   #1  |  Link
Tyestor
Registered User
 
Join Date: Sep 2012
Posts: 12
I don't really understand this technology.

Can someone explain to me how this is possible. These numbers will definitely be off, by the way, they're just examples pulled out of the air.

H264

Resolution: 1920x1080
Bitrate: 3500
Filesize: 1.6GB

H265

Resolution: 1920x1080
Bitrate: 1750
Filesize: 800MB

Am I to assume that these video files would have IDENTICAL "quality"? And if the H265 video's bitrate was doubled (the same as the H264 file) it's, "quality" would be "doubled"? This kind of seems unreal to me. Sorry if I sound skeptical (I kind of am).
Tyestor is offline   Reply With Quote
Old 4th December 2015, 18:58   #2  |  Link
smok3
brontosaurusrex
 
smok3's Avatar
 
Join Date: Oct 2001
Posts: 2,392
Quote:
Originally Posted by Tyestor View Post
Am I to assume that these video files would have IDENTICAL "quality"? And if the H265 video's bitrate was doubled (the same as the H264 file) it's, "quality" would be "doubled"?
Of course not, where did you get that idea?
smok3 is offline   Reply With Quote
Old 4th December 2015, 19:32   #3  |  Link
sneaker_ger
Registered User
 
Join Date: Dec 2002
Posts: 5,565
Quote:
Originally Posted by Tyestor View Post
Am I to assume that these video files would have IDENTICAL "quality"?
That is what the HEVC marketing wants you to believe. In many cases this is not true, though. (But in some cases it may be.)
sneaker_ger is offline   Reply With Quote
Old 4th December 2015, 19:33   #4  |  Link
Tyestor
Registered User
 
Join Date: Sep 2012
Posts: 12
Quote:
Originally Posted by smok3 View Post
Of course not, where did you get that idea?
It's what I read about HEVC. Why do you think I made this thread?

Quote:
Originally Posted by sneaker_ger View Post
That is what the HEVC marketing wants you to believe. In many cases this is not true, though. (But in some cases it may be.)
Thanks for the answer, I knew it couldn't be true.
Tyestor is offline   Reply With Quote
Old 4th December 2015, 20:04   #5  |  Link
x265_Project
Guest
 
Posts: n/a
Quote:
Originally Posted by sneaker_ger View Post
That is what the HEVC marketing wants you to believe. In many cases this is not true, though. (But in some cases it may be.)
When marketers or reporters try to explain HEVC, they try to simplify things to get their point across without creating too much confusion. But like most technical subjects, oversimplifying things can lead to even more confusion. To understand video compression efficiency, it's probably better to use specific examples.

In general, HEVC (H.265) has demonstrated the capability to compress video significantly more efficiently than AVC (H.264). The degree of compression efficiency improvement depends on many things - the video content, the codecs being compared, and the codec settings used.

At sufficiently high bit rates, the difference in visual quality between AVC and HEVC is minimal. Give any video codec enough bits and the video will look great. At low bit rates, the differences can be significant. To objectively measure compression efficiency at different bit rates video codec developers construct a chart called a rate-distortion curve. An example is below...

We conduct a bunch of tests, and we plot the objective quality measurement (usually PSNR or SSIM) against the bit rate for each codec at each test bit rate. When you look at these rate-distortion curves you can see what bit rate you would need with one codec to match the objective quality of the other codec at its bit rate, and in this way you can get a good rough estimate of the difference in compression efficiency at this particular quality level.

Two things to note...
1 - Objective quality measures like PSNR or SSIM will not correlate perfectly with visual quality as perceived by real human subjects. These types of measurements are helpful, but not perfect.
2 - Encoding speed (or the compute power required to achieve a particular speed) is almost always a real factor. Today, if you compare two codecs like x264 and x265 at equivalent encoding speed on a particular hardware platform, x264 compares very well. HEVC is a more complex encoding standard than AVC, and generally it requires more computations to realize its full potential.
  Reply With Quote
Old 5th December 2015, 13:49   #6  |  Link
smok3
brontosaurusrex
 
smok3's Avatar
 
Join Date: Oct 2001
Posts: 2,392
a. x265_Project: Very interesting, so at 1000 kbps, for this specific sample, with this specific settings for each codec (and so on), the PSNR diff between x.264 and x.265 is around 1.5 dB, what can one conclude from that? (It would probably be more interesting to see the 4k video curve example)

b. One could also conclude that the [S]quality[/S] PSNR for x.265 rises much faster than for x.264 (at least the low-bitrate 100-1000 part)? < Just wondering if that is actually meaningful for any specific application.

c. One could also wonder if the curve is actually mostly linear between 1000-2000 kbps?

d. Usability wise, application specific: for example if I would use a x264 or x.265 for intermediate previewing purposes for HD video, i'd still need to use pretty high bitrate, it comes to around 10 megabits (usually i have bunch of grain) for x.264. < So what I'am saying is that when you put both into real world, it gets almost infinitive complex?

Last edited by smok3; 5th December 2015 at 14:20.
smok3 is offline   Reply With Quote
Old 5th December 2015, 19:53   #7  |  Link
x265_Project
Guest
 
Posts: n/a
Quote:
Originally Posted by smok3 View Post
a. x265_Project: Very interesting, so at 1000 kbps, for this specific sample, with this specific settings for each codec (and so on), the PSNR diff between x.264 and x.265 is around 1.5 dB, what can one conclude from that? (It would probably be more interesting to see the 4k video curve example)

b. One could also conclude that the [S]quality[/S] PSNR for x.265 rises much faster than for x.264 (at least the low-bitrate 100-1000 part)? < Just wondering if that is actually meaningful for any specific application.

c. One could also wonder if the curve is actually mostly linear between 1000-2000 kbps?

d. Usability wise, application specific: for example if I would use a x264 or x.265 for intermediate previewing purposes for HD video, i'd still need to use pretty high bitrate, it comes to around 10 megabits (usually i have bunch of grain) for x.264. < So what I'am saying is that when you put both into real world, it gets almost infinitive complex?
Don't read too much into this particular rate-distortion curve. It was just an example picture of an RD curve I found with a Google search. When I have time later I could post some RD curves from more recent tests. You can create your own if you run x264 and x265 at a bunch of different bit rates and measure PSNR or SSIM. But again, the more important thing is the actual visual quality, not the objective quality measurement. It's also important to evaluate visual quality by watching video, not by putting still images (single frames) under a microscope.
  Reply With Quote
Old 7th December 2015, 23:33   #8  |  Link
pandy
Registered User
 
Join Date: Mar 2006
Posts: 1,049
Important for PSNR (and dB in general) is that every 6dB gain means improve quality twice (so provided graphs are kind of correlated with claim reduce bit rate by half or twice better quality with same bitrate).
pandy 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 09:55.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.