Log in

View Full Version : What does "Target quality" exactly mean ?


mike23
30th January 2007, 15:12
In AGK there is an option to specify the "target quality".

Assume my source video (which I want to re-encode) has a bitrate of 1000 kbps and I specify a target quality of 90%, does that mean that the resulting video has a bitrate of (approx) 900 kbps ?

If not: Is it possible to include in future versions a fourth parameter to specify the target bitrate ?

Thank you for the answer
Mike

Sharktooth
31st January 2007, 19:30
target quality is (usually) referred to the average quantizer that will be used.
a target quality of 100% means the encoder will use a average quant of 2. Less quality means a higher quantizer.
Size and bitrate doesnt matter in a quality encoding mode.
if you want to aim for a specific bitrate use the 2-pass mode.

mike23
1st February 2007, 11:32
if you want to aim for a specific bitrate use the 2-pass mode.

Hmm, sorry for this question: But how do I "use the 2-pass mode" ?

I have searched now all possible option in AGK and did not found a "2-pass" mode.

At least in "Advanced Settings" there is no "2-pass" mode
option.

jggimi
1st February 2007, 14:02
Select a specific output size. That's all you need to do.

See the Tutorial included with the software for more assistance.

weaver4
1st February 2007, 16:20
Can someone tell me what Target Quality should I use for XviD quantizer values of: 2.5, 3.0 and 3.5?

Thanks,

weaver4
3rd February 2007, 17:26
Can someone tell me what Target Quality should I use for XviD, single-pass, for quantizer values of: 2.5, 3.0 and 3.5?

I think it is something like this:

Q--->Target quality
2.5-->75%
3.0-->67%
3.5-->60%


Thanks,

jggimi
3rd February 2007, 19:13
@ weaver4:

I have merged your new thread into this one, since it was a reposting of the same question, on the same subject matter.

BigDid
3rd February 2007, 19:14
Can someone tell me what Target Quality should I use for XviD, single-pass, for quantizer values of: 2.5, 3.0 and 3.5?

Hi,

Fire AGK, Make a one pass target quality, abort when running, look at the xvid encoder/target quantizer box and with trial and retry you should get it.
2.5 is 80%
2.99 is 67%
3.51 is 57%
for a short Ntsc test file (divx trailer lastmanstanding)

Did

Edit: this is multithread work :) In the same time I was trying to answer to the other thread by referring to this one I had an invalid thread; I understand now why (threads merged)

jeffy
3rd February 2007, 19:30
Can someone tell me what Target Quality should I use for XviD, single-pass, for quantizer values of: 2.5, 3.0 and 3.5?

I think it is something like this:

Q--->Target quality
2.5-->75%
3.0-->67%
3.5-->60%


Thanks,

AutoGK 2.40
- My experience, if wrong, please post correction.

2.5 = 80 %
2.67 = 75 %
2.99 = 67 %
3.51 = 57 %
4.0 = 50 %

The same: http://forum.doom9.org/showthread.php?p=947621#post947621

weaver4
5th February 2007, 18:08
Thanks, Those are the values that I was looking for.

weaver4
5th February 2007, 20:13
One think that I have yet to understand is that if I do a quantizer of 3.00 in avi.net, staxrip, or AutoMKV they are all roughly the same filesize. Do the same movie in AutoGK at 67% and it is roughly 15 to 20% larger. Same Audio bitrate, same resolution.

Try as I might, I cannot see any quality differences between AutoGK at 67% and avi.net with a quantizer value of 3.

BigDid
5th February 2007, 23:30
... Do the same movie in AutoGK at 67% and it is roughly 15 to 20% larger...
Hi,

The numbers given in this thread are "expected" quantizer; once the 1pass encode is made, maybe you will need to verify the "real" quantizer with the ad-hoc tool (like DRF analyser).

First time we exchanged about this subject (in another thread) the number I had given was the result of a 2pass encode verified with DRFanalyser (mean/average quantizer).

Did

weaver4
6th February 2007, 16:13
I don't think that DRF analyser works anymore with movies that are done with AutoGK, at least mine doesn't. I think it has something to do with the fact that a custom matrix was used in the latest versions of AutoGK

BigDid
6th February 2007, 17:26
I don't think that DRF analyser works anymore with movies that are done with AutoGK, at least mine doesn't.
Hi,

Mine does, example from AmeliePoulain used for my AGK speedtest here:
http://forum.doom9.org/showthread.php?p=949322#post949322
Results:
DivX DRF Analyzer v0.9.5 Report!
File Name: J:\AmelieP\AmeliePtest.avi
FourCC: XVID
Codec: XviD0047
Resolution: [ Width: 640 Height: 272 ]
Frame Rate: 25.000 frames per second
The Video has 174958 frames [ 01:56:38 ]

Average Frame quality is MEDIUM [Average DRF/quantizer is 3.40]
Standard Deviation: Quality is MEDIUM [Std. Deviation is 1.04]
Image Resolution is HIGH

There are NO frame drops ( NO drops is better )
This video may have some frames with VERY POOR Quality!

Recomended Resolution: [576x240] (Target DRF/quantizer=2.8)

Performance Caracteristics:
Macroblocks per frame: 680
The Width is multiple of 32

Kilobits per Second: 1353.02
Kilobits per Frame: 54.12
Kilobits per Macroblock: 0.080
Bits per Pixel: 0.32

Frame Type Statistics :
I Frames: 1.04%
P Frames: 50.41%
B Frames: 48.55%
S Frames: 0.00%
N Frames: 0.00%
(More Advanced Codecs use B and S frames)
Frame Quality Statistics :

DRF=1&2: 40898 23.6%
DRF=3: 43308 25.0%
DRF=4: 60155 34.7%
DRF=5: 24400 14.1%
DRF=6: 987 0.6%
DRF=7: 505 0.3%
DRF=8: 315 0.2%
DRF=9: 719 0.4%
DRF>9: 1853 1.1%
KeyF/DeltaF: 1.05%
KeyDRF<4: 1738
KeyDRF=4: 27
KeyDRF>4: 53

AverageKeyDRF: 2.75
MAXDRF: 11
AverageDRF: 3.40
Deviation: 1.04

I think it has something to do with the fact that a custom matrix was used in the latest versions of AutoGK
AGK uses Jawor 1cd cqm and switch to SharktoothV2 for sharp matrix unless you specify .newsoftmatrix in Agk directory in which case it will use Sharktooth V3ULR for soft matrix.
You can use Mpegmodifier to view the inter/intra numbers and the main xvid cqm are listed here:
http://forum.doom9.org/showthread.php?p=931109#post931109

Did

weaver4
6th February 2007, 21:50
I stand corrected the Latest version of DRFAnalyzer 9.5.1 does work.