Explanation for new switches in x264 - Page 2

Thunderbolt8 · 3rd October 2006, 23:51

ok, thx.

asdfsauce · 3rd October 2006, 23:58

I haven't seen it mentioned in this thread yet, so let me ask. Am I correct in assuming trellis should be turned off for the deadzone switch to have any effect?

akupenguin · 4th October 2006, 01:40

Yes.
Theoretically, the deadzone switches could apply to trellis too by modifying lambda, but that hasn't been implemented yet.

AlexW · 4th October 2006, 04:29

Quote:

Originally Posted by Sharktooth

Can you please also specify the encoding time relation between default values and user set values>

I haven't tested this much yet, but the effect of custom deadzone on encoding speed should be minimal.

Prettz · 4th October 2006, 18:52

Excellent. This sounds like just the option I've been needing to deal with a few very tough scenes in Terminator 2 that weren't responding well to just increased bitrate. I'll start testing it ASAP.

Caroliano · 14th October 2006, 00:47

How deadzone interacts with in-loop deblocking? Is an low value for deblocking needed in order to get good efects from low deadzones values? Or the defaut deblocking (0,0) don't bother?

What's the diference between low deadzone setings and high bitrate CQMs?

Is higher than defaut deadzone values recomended for low-bitrates?

akupenguin · 14th October 2006, 02:54

Deadzone affects the quality-per-qp. Optimal (absolute) deblocking strength depends on quality. Actual deblocking strength (as defined by the standard and set by x264 options) is relative to qp, not relative to quality. So if you change quality-per-qp, then you might want to modify the deblocking setting.
However, the units of the deblocking setting a pretty coarse. So even if you set deadzone all the way down to 0 (not recommended), the compensating change would be to reduce deblock by at most 1 (compared to whatever deblocking you were previously using with the default deadzone).
Conversely, if you increase deadzone all the way up to 32, I expect the optimal deblock setting to change by at most +1.

There is no particular relationship between bitrate and optimal deadzone. Deadzone is already specified relative to the rate-distorion optimal value, so any changes are psychovisual, and not well enough tested yet to give a definitive recommendation.

DryFire · 14th October 2006, 05:03

What would be the difference between using your denoiser of choice as opposed to deadzone?

Speed?

pinkie_1 · 4th February 2007, 13:36

A question to the people with a better insight than mine : how are the --deadzone switches correctly used ?
I've noticed that some apparently odd combinations are quite popular. For instance (4,4) or (6,6,), even (0,0)... (mind you, the first parameter stands for intra, the second for inter)
AFAIK :
1. intra should have a lower value than inter
2. (0,0) shouldn't be used at all
Personally, I tend to use (2,4) or (4,8) when tweaking the --deadzone settings is appropriate.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

3rd October 2006, 23:51	#21 \| Link
Thunderbolt8 Registered User Join Date: Sep 2006 Posts: 2,200	ok, thx.

3rd October 2006, 23:58	#22 \| Link
asdfsauce VEIG Master Join Date: Nov 2005 Location: US/CA Posts: 73	I haven't seen it mentioned in this thread yet, so let me ask. Am I correct in assuming trellis should be turned off for the deadzone switch to have any effect?

4th October 2006, 01:40	#23 \| Link
akupenguin x264 developer Join Date: Sep 2004 Posts: 2,392	Yes. Theoretically, the deadzone switches could apply to trellis too by modifying lambda, but that hasn't been implemented yet.

4th October 2006, 18:52	#25 \| Link
Prettz easily bamboozled user Join Date: Sep 2002 Location: Atlanta Posts: 373	Excellent. This sounds like just the option I've been needing to deal with a few very tough scenes in Terminator 2 that weren't responding well to just increased bitrate. I'll start testing it ASAP.

14th October 2006, 00:47	#26 \| Link
Caroliano Registered User Join Date: Feb 2005 Location: São Paulo, Brazil Posts: 392	How deadzone interacts with in-loop deblocking? Is an low value for deblocking needed in order to get good efects from low deadzones values? Or the defaut deblocking (0,0) don't bother? What's the diference between low deadzone setings and high bitrate CQMs? Is higher than defaut deadzone values recomended for low-bitrates?

14th October 2006, 02:54	#27 \| Link
akupenguin x264 developer Join Date: Sep 2004 Posts: 2,392	Deadzone affects the quality-per-qp. Optimal (absolute) deblocking strength depends on quality. Actual deblocking strength (as defined by the standard and set by x264 options) is relative to qp, not relative to quality. So if you change quality-per-qp, then you might want to modify the deblocking setting. However, the units of the deblocking setting a pretty coarse. So even if you set deadzone all the way down to 0 (not recommended), the compensating change would be to reduce deblock by at most 1 (compared to whatever deblocking you were previously using with the default deadzone). Conversely, if you increase deadzone all the way up to 32, I expect the optimal deblock setting to change by at most +1. There is no particular relationship between bitrate and optimal deadzone. Deadzone is already specified relative to the rate-distorion optimal value, so any changes are psychovisual, and not well enough tested yet to give a definitive recommendation.

14th October 2006, 05:03	#28 \| Link
DryFire Perpetually Lost Join Date: Sep 2005 Posts: 93	What would be the difference between using your denoiser of choice as opposed to deadzone? Speed?

4th February 2007, 13:36	#29 \| Link
pinkie_1 Registered User Join Date: Nov 2006 Posts: 56	A question to the people with a better insight than mine : how are the --deadzone switches correctly used ? I've noticed that some apparently odd combinations are quite popular. For instance (4,4) or (6,6,), even (0,0)... (mind you, the first parameter stands for intra, the second for inter) AFAIK : 1. intra should have a lower value than inter 2. (0,0) shouldn't be used at all Personally, I tend to use (2,4) or (4,8) when tweaking the --deadzone settings is appropriate.