Autoqmatenc

Prodater64 · #1 11-11-2005, 11:59 AM

@Kwag:
I think there is no more legal issues.
Could you reopen the thread?

http://www.kvcd.net/forum/viewtopic.php?p=118494#118494

I think now it really deserves a space here.

Quote:

0.31b officially released :

0.31b - 2005/11/11
* Introduced a new version of QMatOp, it should be faster for bitrates higher than 3750kbps
* Added an number of frames autodetection when last_frame = 0 or not specified. (Introduced for DIKO)
* Here is my first template for DIKO (not tested yet)
* Changed the Q range valueof OPV from a linear one to a logarithmic one...it seems to work better with DVDRB.
* Tweaked the RC for OPV
* Corrected a bug in the PPM writing core function, it was discovered by Danpos from the BDVD Team. Thank you for your tests.
* Corrected a major flaw in the QMatOp engine...Some flat matrices were generated in some cases. Thank you oxa for finding it and running many tests.
* Corrected a bug found by Danpos (BDVD Team) : When the AVS script was producing a Field Based movie, the script refused to open. Thanks again to him !!!!
* Corrected minor bugs
* Removed limitations to the QMatOp engine + Adjusted the core formula (better for low bitrates)
* Added a parameter to change the speed of QMatOp (faster means means less precision) tc_ref_frm, default value 0, above 3 not tested.
* Generated matrices are now scan optimized (pure ZigZag or pure Alternate)
* Matrices are curved smoothed with quadrilinear functions.
* Adjusted the rate control
* Corrected the threshold for Ultra Low Bitrates encodes, it takes into account the resolution.
* Adjusted QMatOp to benefit from the quadrilinear interpolation functions...Faster and better.
* Added a remaining time info
* Corrected a bug found by oxa (blinking squares due to a very low quantization)
* Added a new VAF analyzer to assess more accurately the Qmin/Qmax after the first pass.
* Added a min bitrate correction function
* Spikes should be gone at high bitrates !!!!!
* A stupid error in the ECL parsing was corrected (Thank you oxa for this one)

Also EclCreate is a good GUI to AQE.

http://www.vmesquita.com/forum/index...23482#msg23482

fabrice · #2 11-11-2005, 12:48 PM

Hi,

I'm playimg with this baby, to check underflow & bitrate spike, because it's based on mencoder, but with a self made rate control (like qenc).

Till now, I've been playing with q mode, but it's difficult to get a given filesize...

More news from my part on monday!

salu2
Fabrice

Prodater64 · #3 11-11-2005, 01:28 PM

Quote:

Originally Posted by fabrice

Hi,

I'm playimg with this baby, to check underflow & bitrate spike, because it's based on mencoder, but with a self made rate control (like qenc).

Till now, I've been playing with q mode, but it's difficult to get a given filesize...

More news from my part on monday!

salu2
Fabrice

I don't know Q mode, don't tested yet, but in 2PVBR it is almost so exact as HC Encoder is with no spikes over compliant!!!
Today was realeased last beta.
http://sapstar.vmesquita.com

Prodater64 · #4 11-13-2005, 07:26 AM

Quote:

Originally Posted by fabrice

Hi,
Till now, I've been playing with q mode, but it's difficult to get a given filesize...
Fabrice

Last version is more accurate regarding final size.
As you maybe are doing prediction manually, and it is a lot of work, you can use it with DIKO-ECLCREATE-AQE set.

http://www.vmesquita.com/forum/index.php?topic=4898.0

@Kwag: Did you miss my first post?

kwag · #5 11-13-2005, 10:08 AM

Quote:

Originally Posted by Prodater64

@Kwag: Did you miss my first post?

Yes I did

Damn power fail and I lost the browser cookies

I'll check into this later today.

-Karl

fabrice · #6 11-14-2005, 12:29 AM

Hi again,

I've been playing a bit this week end with OPV, and I ended in encoding a movie at 704x576, with an average bitrate of 1388 (manual prediction!

).

With HC, this movie come almost perfect, but with AQE, there is some very ugly artifact is some very bitrate demanding scene (fire, ghosts, ...).
The final size is almost on target, but BV tell me that in this ugly scenes, the Q rise to 15, 29 during more almost 1 minute... With this bitrate, I get spike of 6018 (max bitrate is 8000)

I'll try today with a 2 pass encoding, to see the result.

Cu,
Fabrice

kwag · #7 11-14-2005, 12:38 AM

Well, I was about to re-open that thread, but after reading Fabrice's post, I'm going to keep it locked until further tests.

-Karl

Prodater64 · #8 11-14-2005, 01:37 AM

Quote:

Originally Posted by kwag

Well, I was about to re-open that thread, but after reading Fabrice's post, I'm going to keep it locked until further tests.

-Karl

I can say you that first test with OPV mode gives me -0.02 % difference.
Very good quality.

Dialhot · #9 11-14-2005, 04:07 AM

Quote:

Originally Posted by kwag

Well, I was about to re-open that thread, but after reading Fabrice's post, I'm going to keep it locked until further tests.

-Karl

The thread was locked for legal issues.
We can close it again later if the encoder has big compliance issues, like we did for mencoder. But meanwhile, I don't see a reason to let it close if other people than Luis en Fabrice want to test it.

My 2 cents.

Prodater64 · #10 11-14-2005, 05:45 AM

Quote:

Originally Posted by fabrice

Hi again,

I've been playing a bit this week end with OPV, and I ended in encoding a movie at 704x576, with an average bitrate of 1388 (manual prediction!

).

With HC, this movie come almost perfect, but with AQE, there is some very ugly artifact is some very bitrate demanding scene (fire, ghosts, ...).
The final size is almost on target, but BV tell me that in this ugly scenes, the Q rise to 15, 29 during more almost 1 minute... With this bitrate, I get spike of 6018 (max bitrate is 8000)

I'll try today with a 2 pass encoding, to see the result.

Cu,
Fabrice

1 - Do you tried last AQE version?
2 - You should use QMatOp = ON, as SAPSTAR worked tunning it and at the moment didn't tweak the encoder with QMatOp = OFF.
3 - If you want to avoid working with manual prediction, you can try this method i descripted: http://www.vmesquita.com/forum/index.php?topic=4898.0 , it accelerates tests.

Prodater64 · #11 11-14-2005, 06:11 AM

Quote:

Originally Posted by Dialhot

The thread was locked for legal issues.
We can close it again later if the encoder has big compliance issues, like we did for mencoder. But meanwhile, I don't see a reason to let it close if other people than Luis en Fabrice want to test it.

My 2 cents.

Thanks.
Indeed, it is why i opened this thread.