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 5th May 2019, 13:33   #1  |  Link
excellentswordfight
Lost my old account :(
 
Join Date: Jul 2017
Posts: 322
Is no-strong-intra-smoothing really beneficial?

I've been doing some tests with common settings that I see people use arround here for detail retention, and since there is no good tune preset for this, it seems to differ a bit what settings should be used as an substitute for an tune film preset.

For my medium bitrate test (1080p @ 6Mbps) I found that no-strong-intra-smoothing actually only displayed negative effects. So my question is why I see a lot of people using it. Have you done tests or is this somthing that just sounds like it should be befinitial so it's just assumed that it it is? Or maybe its only better at higher bitrate?



Default: A clear loss in detail, some areas are blurry.

No-sao: Most detail restored, some areas are still blurry

No-sao, deblock -1,-1: Most detail restored, most of the blurry areas are gone.

No-sao, deblock -1,-1, no-strong-intra-smoothing: Most detail restored, blurry areas from the default is back. While not visable in sample, some areas display blocky tendencies simulair to an comparable x264 encode.

Tests are done with Tears of steal uncompressed 4k source downscaled to 1080p encoded with x265 v3.0+14 2pass preset slow. 6Mbps equals crf19 for this sample.

Last edited by excellentswordfight; 5th May 2019 at 16:15.
excellentswordfight is offline   Reply With Quote
Old 5th May 2019, 15:01   #2  |  Link
Nico8583
Registered User
 
Join Date: Jan 2010
Location: France
Posts: 851
Very interesting topic, I follow it. I will do some tests.
Nico8583 is offline   Reply With Quote
Old 5th May 2019, 17:18   #3  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Finland
Posts: 5,718
I think the non-negative version of the parameter name itself makes people shudder, so they switch it off (me included). Your tests look very interesting indeed. Was the average bitrate almost the same even with the blurry areas? I was just wondering where the bits go if it is..

EDIT: Just for the heck of it, I did a quick search and found an interesting paper saying "..The process was developed to remove some blocking and contouring artifacts visible on extremely smooth image areas.."
__________________
And if the band you're in starts playing different tunes
I'll see you on the dark side of the Moon...

Last edited by Boulder; 5th May 2019 at 17:25.
Boulder is offline   Reply With Quote
Old 5th May 2019, 18:42   #4  |  Link
Selur
Registered User
 
Selur's Avatar
 
Join Date: Oct 2001
Location: Germany
Posts: 7,259
Quote:
"..The process was developed to remove some blocking and contouring artifacts visible on extremely smooth image areas.."
so it might be a good option for cartoons
__________________
Hybrid here in the forum, homepage
Selur is offline   Reply With Quote
Old 5th May 2019, 18:57   #5  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Finland
Posts: 5,718
Based on those images, extremely smooth is really not that smooth but more like any quite nicely compressible flat-like area. At least if the effect is what it is doing there..
__________________
And if the band you're in starts playing different tunes
I'll see you on the dark side of the Moon...
Boulder is offline   Reply With Quote
Old 5th May 2019, 22:39   #6  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,750
Quote:
Originally Posted by Nico8583 View Post
Very interesting topic, I follow it. I will do some tests.
The real question is quality @ bitrate? If it adds detail AND bitrate it isn't clear if it is a better deal than using SAO at a lower CRF. Comparing in 2-pass VBR mode would allow apples-to-apples.

I would expect SAO to help more as bitrates go down and CRF goes up.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 6th May 2019, 06:03   #7  |  Link
StvG
Registered User
 
Join Date: Jul 2018
Posts: 447
@excellentswordfight, what's the source bitrate?
StvG is offline   Reply With Quote
Old 6th May 2019, 08:12   #8  |  Link
jd17
Registered User
 
Join Date: Jun 2017
Posts: 89
I always wondered myself (why people seem to disable strong-intra-smoothing), but was too lazy to complain about it.
My (simple) tests showed the same as yours @excellentswordfight, which is why I always kept strong-intra-smoothing.

Thanks for including deblock in your test series!
I was always quite happy with the default 0,0 - but -1,-1 is cheap (no real speed or bitrate penalty as far as I remember) so I might add it to my default settings.

Quote:
Originally Posted by benwaggoner View Post
The real question is quality @ bitrate? If it adds detail AND bitrate it isn't clear if it is a better deal than using SAO at a lower CRF. Comparing in 2-pass VBR mode would allow apples-to-apples.

I would expect SAO to help more as bitrates go down and CRF goes up.
The lowest bitrates I tested were around 1000kbit/s and still, sao never helped in my eyes.
Also, no-sao does not slow things down or add significant bitrate with reasonable CRFs in my experience.

I always come to the same conclusion, it is a loss/loss to me and I still don't understand how it is default for almost every preset.

It is the one "switch" that changes x265 from good to amazing imho.
jd17 is offline   Reply With Quote
Old 6th May 2019, 08:34   #9  |  Link
excellentswordfight
Lost my old account :(
 
Join Date: Jul 2017
Posts: 322
Quote:
Originally Posted by Boulder View Post
Based on those images, extremely smooth is really not that smooth but more like any quite nicely compressible flat-like area. At least if the effect is what it is doing there..
The question is if the lack off strong-intra-smoothing is directly related to the defects in this picture, or if it saves that much bitrate elsewere so that bits can be spent on details in areas that are otherwise blury. Eitherway, I can't really find any parts that looks better with it disabled, so it's not like it causes obvious smoothing were it shouldn't.

Quote:
Originally Posted by benwaggoner View Post
The real question is quality @ bitrate? If it adds detail AND bitrate it isn't clear if it is a better deal than using SAO at a lower CRF. Comparing in 2-pass VBR mode would allow apples-to-apples.

I would expect SAO to help more as bitrates go down and CRF goes up.
All tests are 2pass VBR, and the question is not regarding SAO but strong intra smoothing. But in terms of SAO, disabling it have never caused any side effects for me outside animation and very low bitrate material. It doesnt effect encoding speed, or increase bitrate @ given CRF level.

Quote:
Originally Posted by StvG View Post
@excellentswordfight, what's the source bitrate?
As stated, uncompressed (not really true though, its re-compressed to FFV1, so actually lossless not uncompressed).

Last edited by excellentswordfight; 6th May 2019 at 10:55.
excellentswordfight is offline   Reply With Quote
Old 8th May 2019, 12:30   #10  |  Link
Gser
Registered User
 
Join Date: Apr 2008
Posts: 418
The comparison would have been better with a picture of the source frame and not all just compressed frames. From what I remember from my testing, disabling intra smoothing helped retain grain and other similar fine details like skin texture.

In addition to other complaints about x265, I still can't comprehend why deblock 0:0 is default like in x264, this singular setting is the biggest reason a lot of encodes, including broadcast, look so blurry. I would rather have a little bit of macroblocking than the picture turn into complete pudding. Yeah yeah you encode blu-rays as your profession at a gigazillion bytes per second, that's only a very small fraction of what video is compressed. Personally I use -2:-3 at crf 22-24 but even a -1:-1 already gives a massive improvement.

Last edited by Gser; 8th May 2019 at 12:40.
Gser is offline   Reply With Quote
Old 8th May 2019, 14:47   #11  |  Link
excellentswordfight
Lost my old account :(
 
Join Date: Jul 2017
Posts: 322
Quote:
Originally Posted by Gser View Post
The comparison would have been better with a picture of the source frame and not all just compressed frames. From what I remember from my testing, disabling intra smoothing helped retain grain and other similar fine details like skin texture.

In addition to other complaints about x265, I still can't comprehend why deblock 0:0 is default like in x264, this singular setting is the biggest reason a lot of encodes, including broadcast, look so blurry. I would rather have a little bit of macroblocking than the picture turn into complete pudding. Yeah yeah you encode blu-rays as your profession at a gigazillion bytes per second, that's only a very small fraction of what video is compressed. Personally I use -2:-3 at crf 22-24 but even a -1:-1 already gives a massive improvement.
Here is the lossless and the 'slow no-sao deblock' full frame versions. I think the results are very impressive for 6Mbps!




Tried another frame as well, this time a "harder" one. It seems like turning off strong intra smoothing hits efficiency pretty hard, in this frame it creates some heavy blocking.

(from left to right: lossless, 'slow no-sao deblock -1,-1', 'slow no-sao deblock -1,-1 no-strong-intra-smoothing'

Last edited by excellentswordfight; 8th May 2019 at 14:52.
excellentswordfight is offline   Reply With Quote
Old 8th May 2019, 19:39   #12  |  Link
Forteen88
Herr
 
Join Date: Apr 2009
Location: North Europe
Posts: 556
Quote:
Originally Posted by Gser View Post
In addition to other complaints about x265, I still can't comprehend why deblock 0:0 is default like in x264, this singular setting is the biggest reason a lot of encodes, including broadcast, look so blurry. I would rather have a little bit of macroblocking than the picture turn into complete pudding. Yeah yeah you encode blu-rays as your profession at a gigazillion bytes per second, that's only a very small fraction of what video is compressed. Personally I use -2:-3 at crf 22-24 but even a -1:-1 already gives a massive improvement.
Probably because having a high deblock-value increases compression (according to b.waggoner here). But yeah, for transparent encodes, having a lower deblock-value (to a point) is great.
Forteen88 is offline   Reply With Quote
Old 8th May 2019, 20:32   #13  |  Link
jd17
Registered User
 
Join Date: Jun 2017
Posts: 89
Quote:
Originally Posted by Gser View Post
I still can't comprehend why deblock 0:0 is default like in x264, this singular setting is the biggest reason a lot of encodes, including broadcast, look so blurry. I would rather have a little bit of macroblocking than the picture turn into complete pudding. Yeah yeah you encode blu-rays as your profession at a gigazillion bytes per second, that's only a very small fraction of what video is compressed. Personally I use -2:-3 at crf 22-24 but even a -1:-1 already gives a massive improvement.
At least judging from my own tests, I do not agree with this.
It sounds very exaggerated to me.

"Blurry", "pudding" and "massive improvement" for deblock 0:0 vs. -1:-1?
Do you have some example frames that show this?
At what bitrates?
I would sign the statement for sao vs. no-sao, but the difference for deblock is quite subtle in my eyes.
jd17 is offline   Reply With Quote
Old 8th May 2019, 20:39   #14  |  Link
jd17
Registered User
 
Join Date: Jun 2017
Posts: 89
Quote:
Originally Posted by excellentswordfight View Post
Here is the lossless and the 'slow no-sao deblock' full frame versions. I think the results are very impressive for 6Mbps!


The result does look good, however I assume the encode is 10bit and the lossless source is 8bit?
The screenshot shows a difference in brightness (and more?), which I found to be common when comparing 8bit to 10bit videos.

Quote:
It seems like turning off strong intra smoothing hits efficiency pretty hard, in this frame it creates some heavy blocking.
This is in line with the types of artifacts I remember from testing no-strong-intra...
Thanks for sharing.

Last edited by jd17; 8th May 2019 at 20:42.
jd17 is offline   Reply With Quote
Old 9th May 2019, 13:26   #15  |  Link
Gser
Registered User
 
Join Date: Apr 2008
Posts: 418
Quote:
Originally Posted by jd17 View Post
The result does look good, however I assume the encode is 10bit and the lossless source is 8bit?
The screenshot shows a difference in brightness (and more?), which I found to be common when comparing 8bit to 10bit videos.


This is in line with the types of artifacts I remember from testing no-strong-intra...
Thanks for sharing.
Yeah I have noticed this brightness difference as well. Strange.
And perhaps I just have perceived these artifacts as increased sharpness.
Gser 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 20:44.


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