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 28th May 2020, 20:29   #7621  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,988
3.4 has been tagged for release!

https://bitbucket.org/multicoreware/...ch/Release_3.4

Release notes - v3.4
-----------------------------

New features
------------------
• Edge-aware quadtree partitioning to terminate CU depth recursion based on edge information. --rskip 2 enables the feature and --rskip-edge-threshold <0..100> denotes the minimum expected edge-density percentage within the CU, below which the recursion is skipped. Experimental feature.
• Application-level feature --abr-ladder for automating ABR ladder generation. Shows ~65% savings in the over-all turn-around time required for the generation of a typical HLS ladder in Intel(R) Xeon(R) Platinum 8280 CPU @ 2.70GHz over a sequential ABR-ladder generation approach that leverages save-load architecture.
Enhancements to existing features
----------------------------------------------
• Improved 2-pass rate-control algorithm. The savings in the bitrate is ~1.72% with visual improvement in quality in the initial 1-2 secs.
Encoder enhancements
--------------------------------
• Faster ARM64 encodes enabled by ASM contributions from Huawei. The speed-up over no-asm version for 1080p encodes @ medium preset is ~15% in a 16 core H/W.
• Strict VBV conformance in zone encoding.
Bug fixes
------------
• Multi-pass encode failures with --frame-dup.
• Corrupted bitstreams with --hist-scenecut when input depth and internal bit-depth differ.
• Incorrect analysis propagation in multi-level save-load architecture.
• Failure in detecting NUMA packages installed in non-standard directories.
Blue_MiSfit is offline   Reply With Quote
Old 31st May 2020, 09:17   #7622  |  Link
Patman
Registered User
 
Patman's Avatar
 
Join Date: Jan 2015
Posts: 286
I've updated my x265 builds. x265-3.4+1 is built from the stable branch (stable release) and x265-3.4+6 is built from the master branch. Both builds are 64-bit multilib windows binaries.
__________________
Tools for StaxRip | x264 - x265

Last edited by Patman; 31st May 2020 at 12:25.
Patman is offline   Reply With Quote
Old 31st May 2020, 14:53   #7623  |  Link
alexantr
Registered User
 
alexantr's Avatar
 
Join Date: Nov 2019
Location: Belarus
Posts: 5
Who try --rskip 2? Did you catch rare glitches like that
Attached Images
  
alexantr is offline   Reply With Quote
Old 31st May 2020, 23:22   #7624  |  Link
Sagittaire
Testeur de codecs
 
Sagittaire's Avatar
 
Join Date: May 2003
Location: France
Posts: 2,484
Quote:
Originally Posted by Blue_MiSfit View Post
3.4 has been tagged for release!

https://bitbucket.org/multicoreware/...ch/Release_3.4

Release notes - v3.4
-----------------------------

New features
------------------
• Edge-aware quadtree partitioning to terminate CU depth recursion based on edge information. --rskip 2 enables the feature and --rskip-edge-threshold <0..100> denotes the minimum expected edge-density percentage within the CU, below which the recursion is skipped. Experimental feature.
• Application-level feature --abr-ladder for automating ABR ladder generation. Shows ~65% savings in the over-all turn-around time required for the generation of a typical HLS ladder in Intel(R) Xeon(R) Platinum 8280 CPU @ 2.70GHz over a sequential ABR-ladder generation approach that leverages save-load architecture.
Enhancements to existing features
----------------------------------------------
• Improved 2-pass rate-control algorithm. The savings in the bitrate is ~1.72% with visual improvement in quality in the initial 1-2 secs.
Encoder enhancements
--------------------------------
• Faster ARM64 encodes enabled by ASM contributions from Huawei. The speed-up over no-asm version for 1080p encodes @ medium preset is ~15% in a 16 core H/W.
• Strict VBV conformance in zone encoding.
Bug fixes
------------
• Multi-pass encode failures with --frame-dup.
• Corrupted bitstreams with --hist-scenecut when input depth and internal bit-depth differ.
• Incorrect analysis propagation in multi-level save-load architecture.
• Failure in detecting NUMA packages installed in non-standard directories.
well now Npass encoding seem work like crf mode: It is not a progess. For me the quantizer ratio between pframe and bframe is too high by default.
__________________
Le Sagittaire ... ;-)

1- Ateme AVC or x264
2- VP7 or RV10 only for anime
3- XviD, DivX or WMV9
Sagittaire is offline   Reply With Quote
Old 6th June 2020, 18:34   #7625  |  Link
burnix
Registered User
 
Join Date: Jan 2003
Posts: 49
Patman. Can you build an ffmpeg x64 with it ????

Thanks
burnix is offline   Reply With Quote
Old 6th June 2020, 21:44   #7626  |  Link
Patman
Registered User
 
Patman's Avatar
 
Join Date: Jan 2015
Posts: 286
Quote:
Originally Posted by burnix View Post
Patman. Can you build an ffmpeg x64 with it ????

Thanks
Here
__________________
Tools for StaxRip | x264 - x265
Patman is offline   Reply With Quote
Old 7th June 2020, 08:35   #7627  |  Link
burnix
Registered User
 
Join Date: Jan 2003
Posts: 49
You're the best.
I'm going to try now, and after i'm going to sell on internet all crappy computer and server to make a real decent computer and work on ffmpeg compil.

Thanks man.
burnix is offline   Reply With Quote
Old 8th June 2020, 20:08   #7628  |  Link
K.i.N.G
Registered User
 
Join Date: Aug 2009
Posts: 90
I have been stuggling for months with ugly distortions (smearing that kinda looks like motion blur with banding) in fast moving scenes and finally found out that x265's CUTree is to blame.
Turning it off solves it, but the bitrate blows up when trying to keep the same overal quality in other scenes.

Is there any way to make CUTree 'less agressive' or set it to 50% for example or some threshold parameter?

Last edited by K.i.N.G; 8th June 2020 at 22:49.
K.i.N.G is offline   Reply With Quote
Old 8th June 2020, 21:44   #7629  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,988
try a lower ctu value maybe? What resolution + bitrate are you targeting?
Blue_MiSfit is offline   Reply With Quote
Old 8th June 2020, 22:48   #7630  |  Link
K.i.N.G
Registered User
 
Join Date: Aug 2009
Posts: 90
Quote:
Originally Posted by Blue_MiSfit View Post
try a lower ctu value maybe? What resolution + bitrate are you targeting?
4K Bitrate depends on source, but the current footage is decent quality with little grain (which i'd like to keep).
For this footage, lets say arround 15K since its for streaming.

I restarted the encode with --ctu 32 (instead of --ctu 64)

Last edited by K.i.N.G; 9th June 2020 at 00:15.
K.i.N.G is offline   Reply With Quote
Old 9th June 2020, 19:37   #7631  |  Link
K.i.N.G
Registered User
 
Join Date: Aug 2009
Posts: 90
Lowering the CTU size to 32 (--ctu 32) doesnt really seem to help
K.i.N.G is offline   Reply With Quote
Old 10th June 2020, 05:15   #7632  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Finland
Posts: 5,718
By your description, it looks like the infamous onion artifacts. Have you tried --rd 6 with --rd-refine?
__________________
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 10th June 2020, 06:28   #7633  |  Link
poisondeathray
Registered User
 
Join Date: Sep 2007
Posts: 5,345
Does increasing qcomp reduce the cutree effect? Similar to how increasing qcomp reduced mbtree effect in x264 ?
poisondeathray is offline   Reply With Quote
Old 10th June 2020, 11:55   #7634  |  Link
K.i.N.G
Registered User
 
Join Date: Aug 2009
Posts: 90
Quote:
Originally Posted by Boulder View Post
By your description, it looks like the infamous onion artifacts. Have you tried --rd 6 with --rd-refine?
I have not.
I will make sure to try those out next after this encode, which will run for a few days.

Quote:
Originally Posted by poisondeathray View Post
Does increasing qcomp reduce the cutree effect? Similar to how increasing qcomp reduced mbtree effect in x264 ?
I can confirm it does.
Normally I raise the qcomp to something like 0.70-0.75 and raise the CRF by ±0.5 which 'fixes' it to more acceptable levels.

Though... I always wondered if there was another (more efficient) way to control CUTree's strength/agressiveness more directly... (if I understand correctly qcomp controls a lot more than just things that are influenced by cutree). Hence my question here.

Last edited by K.i.N.G; 10th June 2020 at 12:37.
K.i.N.G is offline   Reply With Quote
Old 10th June 2020, 16:06   #7635  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Finland
Posts: 5,718
Quote:
Originally Posted by K.i.N.G View Post
I have not.
I will make sure to try those out next after this encode, which will run for a few days.
And then it will run even slower
Quote:
Though... I always wondered if there was another (more efficient) way to control CUTree's strength/agressiveness more directly... (if I understand correctly qcomp controls a lot more than just things that are influenced by cutree). Hence my question here.
I wouldn't be surprised if aq-mode also played a part here as in my opinion they are related, and the default mode 2 is something I will not use based on my tests. I'm still using mode 1 with strength 0.8-1.0 depending on the source. Higher for cleaner sources.
__________________
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 10th June 2020, 17:16   #7636  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,750
Quote:
Originally Posted by Boulder View Post
I wouldn't be surprised if aq-mode also played a part here as in my opinion they are related, and the default mode 2 is something I will not use based on my tests. I'm still using mode 1 with strength 0.8-1.0 depending on the source. Higher for cleaner sources.
I note from the x265 dev list that an --auto-aq mode is in development. Maybe we'll be able to stop sweating aq-mode someday.

> This patch does the following:
> 1. Automatically decides the AQ Mode for each frame, using its scene
> statistics, such as luma intensity and edge density.
> 2. Add option "--auto-aq" to enable auto detection of AQ Mode per frame.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 10th June 2020, 23:20   #7637  |  Link
fauxreaper
Registered User
 
Join Date: Oct 2014
Posts: 23
CUTree granularity depends on CU depth recursion. --rskip 0 or 2 reduces onion effect. More early exits from CU recursion = bigger CUs = higher chances of wrong CU propagation and artifacts in motion.

x265 RDO loves to skip blocks, and CUTree artifacts are a consequence of this. Decreasing max-merge also helps, reducing artifacts and skipped blocks. Sometimes even setting bframes=0 decrease artifacts in motion.

Last edited by fauxreaper; 10th June 2020 at 23:29.
fauxreaper is offline   Reply With Quote
Old 11th June 2020, 06:29   #7638  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Finland
Posts: 5,718
Quote:
Originally Posted by fauxreaper View Post
Decreasing max-merge also helps, reducing artifacts and skipped blocks.
I've never understood why max-merge gets increased in the slower and supposedly higher quality profiles. I always have it at '2'.
__________________
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 11th June 2020, 09:42   #7639  |  Link
Boulder
Pig on the wing
 
Boulder's Avatar
 
Join Date: Mar 2002
Location: Finland
Posts: 5,718
Quote:
Originally Posted by fauxreaper View Post
CUTree granularity depends on CU depth recursion. --rskip 0 or 2 reduces onion effect. More early exits from CU recursion = bigger CUs = higher chances of wrong CU propagation and artifacts in motion.

x265 RDO loves to skip blocks
I did a little test with a UHD source, downscaled to 1440p. Interesting results, and easy to see that skips and merges are happening more with the default rskip setting compared to mode 2. I didn't do any visual comparions to find any onion artifacts yet.

Code:
	                rskip 1	rskip 2
 Intra 64x64 DC	        0,00 %	0,00 %
 Intra 64x64 Planar	0,00 %	0,00 %
 Intra 64x64 Ang	0,00 %	0,00 %
 Intra 32x32 DC	        2,03 %	1,24 %
 Intra 32x32 Planar	1,21 %	0,79 %
 Intra 32x32 Ang	9,08 %	3,61 %
 Intra 16x16 DC	        0,92 %	0,36 %
 Intra 16x16 Planar	0,43 %	0,15 %
 Intra 16x16 Ang	8,00 %	2,46 %
 Intra 8x8 DC	        0,25 %	0,14 %
 Intra 8x8 Planar	0,06 %	0,04 %
 Intra 8x8 Ang	        3,75 %	1,61 %
 4x4	                0,45 %	0,44 %
 Inter 64x64	        6,50 %	49,71 %
 Inter 64x64 (Rect)	0,29 %	15,61 %
 Inter 32x32	        19,90 %	6,50 %
 Inter 32x32 (Rect)	1,05 %	0,91 %
 Inter 16x16	        16,78 %	4,27 %
 Inter 16x16 (Rect)	0,91 %	0,05 %
 Inter 8x8	        6,65 %	2,21 %
 Inter 8x8 (Rect)	0,83 %	0,30 %
 Skip 64x64	        0,00 %	0,00 %
 Skip 32x32	        0,81 %	0,24 %
 Skip 16x16	        3,73 %	1,04 %
 Skip 8x8    	        4,27 %	1,91 %
 Merge 64x64	        0,27 %	1,63 %
 Merge 32x32	        2,49 %	0,82 %
 Merge 16x16	        4,86 %	1,67 %
 Merge 8x8	        4,47 %	2,27 %
__________________
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 12th June 2020, 05:20   #7640  |  Link
Greenhorn
Registered User
 
Join Date: Apr 2018
Posts: 61
the comparatively enormous number of 64x64 blocks with rskip 2 to me, though-- from a total of 6.79% (normal + rect) to 65.33% (normal + rect)!

Is there any way to get CU-type in a more readable format than setting csv-log-level=1 and viewing the generated file in an Excel/CSV viewer? Would greatly help with some comparison's I've been doing for myself.
Greenhorn 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 12:40.


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