Quantcast FFmpeg: FFvfw Video Codec - Page 16 - digitalFAQ.com Forums [Archives]
  #301  
02-03-2004, 05:16 PM
incredible incredible is offline
Free Member
 
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via ICQ to incredible
BTW: At all my last samples I did NOT add noise! just the script and their diff. resizers acording to final resolution!

Here now 480x576 at 1164CBR and again the better quantize value of 2!
So its not FIX by this:"yes" the screwing on the CBR Bitrate does compensate a bit. But you have to screw a lot and thats risky.

As you see Kwag I choosed the same quantizer as in case of 352x576 but had to lower the CBR to archive almost the same final size of these 1% at 480x576 compared to the 352x576 sample above.

Final size on 1% is 6.464kbit

(as I understood right Kwag, the VBV 20 below in the screen stand for a real double value like VBV40 in TmpgENc??)

Reply With Quote
Someday, 12:01 PM
admin's Avatar
Site Staff / Ad Manager
 
Join Date: Dec 2002
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
  #302  
02-03-2004, 05:25 PM
Anerboda Anerboda is offline
Free Member
 
Join Date: Mar 2003
Location: Denmark
Posts: 132
Thanks: 0
Thanked 0 Times in 0 Posts
So, I got to get this straight: do we have to change both the min and max Quantizer to be the same value or can we have different values in the min and max settings??

Anerboda
Reply With Quote
  #303  
02-03-2004, 05:26 PM
incredible incredible is offline
Free Member
 
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via ICQ to incredible
I always have quantize max to 25 or even 30 and I ONLY change at the min quantizer from 2 to 3 depends on the final wanted filesize and resolution!

The max value in all my cases did cause nothing as the CBR encode refers to the min Quatize value. You can se that in the FFvfw Graph which you can open during encoding below right at your desctop left hand to the little watch!
Reply With Quote
  #304  
02-03-2004, 05:29 PM
Anerboda Anerboda is offline
Free Member
 
Join Date: Mar 2003
Location: Denmark
Posts: 132
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by incredible
I always have quantize max to 25 or even 30 and I ONLY change at the min quantizer from 2 to 3 depends on the final wanted filesize and resolution!

The max value in all my cases did cause nothing as the CBR encode refers to the min Quatize value. You can se that in the FFvfw Graph which you can open during encoding below right at your desctop left hand to the little watch!
But then the Q-level won't be linear, but perhaps it doesn't have to be??
Reply With Quote
  #305  
02-03-2004, 05:41 PM
incredible incredible is offline
Free Member
 
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via ICQ to incredible
At my KDVD sample 704x576 where I set that high bitrate the Q Level was Linear, but in case of OneCD encodings it isn't as I lower the CBR Bitrate at least to get the right VBV.

Linear Q curve is luxury and means nothing, you just should keep the Q curve about 3 or better below. And thats only a numeral value, only your eyes will tell you the truth.

DO open a TmpgENc CQ encoding and there also even at high CQ Values the Q curve won't be linear.

On the other hand as I said that often: Bitrateviewer and also its Q curve is not a garant for real viewable quality!

Let me quote KIKA from his GOP DCT/ Matrix Guide:

Quote:
Originally Posted by Kika
The force of expression of the q-level with the popular Bitrateviewer is to be enjoyed with caution. Because a high level means only that a block contained many same values after quantization, and which also could mean a result of a perfect interaction of quantization and RLE coding!!! And that again means that in such a case highly one quantized, the image quality however on the highest possible level was! Since however hardly someone really understands, what has it with the q-level actually on itself, I leave rather the fingers of interpretation attempts.
(Googles translation engine just used)

Means: If You encode crap and after reassembling the stream when the player again uses the matrix to descramble and the SAME crap comes out, the result is a 1:1 copy of that crap = "....could mean a result of a perfect interaction of quantization and RLE coding" and means a good Q value in Bitrateviewer. (If I understood KIKA in his guide right)
Reply With Quote
  #306  
02-03-2004, 05:48 PM
Anerboda Anerboda is offline
Free Member
 
Join Date: Mar 2003
Location: Denmark
Posts: 132
Thanks: 0
Thanked 0 Times in 0 Posts
But do you gain anything by not keeping the same min and max values?
Reply With Quote
  #307  
02-03-2004, 05:51 PM
incredible incredible is offline
Free Member
 
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via ICQ to incredible
These are my first encodings on CBR mode in ffvfw and so as I said I do not recommend anything It was just set and I did let it run and as my logic told me that only the min value of quantize makes sense to modify (I figured that out in the ffvfw graph!), so now that was it what I actually did
Reply With Quote
  #308  
02-03-2004, 05:54 PM
Anerboda Anerboda is offline
Free Member
 
Join Date: Mar 2003
Location: Denmark
Posts: 132
Thanks: 0
Thanked 0 Times in 0 Posts
Ok, thanks.

I guess more testing will be done, I don't think we're finished yet

Anerboda
Reply With Quote
  #309  
02-03-2004, 05:58 PM
incredible incredible is offline
Free Member
 
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via ICQ to incredible
We just just figured out "how to drive that car" but we still don't now which type of gas is the best and how fast it could drive then!
Reply With Quote
  #310  
02-03-2004, 06:00 PM
kwag kwag is offline
Free Member
 
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Anerboda
I don't think we're finished yet
I'm not
I'm trying to find a way to encode with 9 bits precision, but that's probably impossible at this moment

-kwag
Reply With Quote
  #311  
02-03-2004, 06:02 PM
kwag kwag is offline
Free Member
 
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by incredible
We just just figured out "how to drive that car" but we still don't now which type of gas is the best and how fast it could drive then!
But we've come from a "Go Kart" to a Porche in a matter of hours


-kwag
Reply With Quote
  #312  
02-03-2004, 06:03 PM
Hydeus Hydeus is offline
Free Member
 
Join Date: Dec 2003
Location: Omicron Persei 8
Posts: 322
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by kwag
I'm trying to find a way to encode with 9 bits precision, but that's probably impossible at this moment
And what will be benefit of this?? Yes, yes .. I'm newbie
__________________
Go for SECAM =)
Reply With Quote
  #313  
02-03-2004, 06:12 PM
kwag kwag is offline
Free Member
 
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Hydeus
Quote:
Originally Posted by kwag
I'm trying to find a way to encode with 9 bits precision, but that's probably impossible at this moment
And what will be benefit of this?? Yes, yes .. I'm newbie
Higher quality on bitrates above ~4Mbps.

-kwag
Reply With Quote
  #314  
02-03-2004, 06:43 PM
incredible incredible is offline
Free Member
 
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via ICQ to incredible
Next Step a full action flashing long Movie at 480x576

"The Core" 130mins!, ... very hard to get good result:

Min Quantizers I=3, P=3, B=4, 84 Frames within a GOP! CBR 1165, 1B Frame (yes it decreased filesize), noise added via ffvfw.

And a little extra Unfilter(-8,-8 ) below the MA Command in my script.
... that little unsharpen the movie does make wonder according to an acceptable quantisation on THAT kind of movie.
Merge/Chroma/Luma Blur would do it also but I don't had that command now in mind and I was just too lazy to look in the orig MA )))

We're still at mpeg2 one CDencoding

690kbit AVG Bitrate!!

Look at the Q curve, it seems the encoder after some Frames does get calm


Reply With Quote
  #315  
02-04-2004, 05:41 AM
Hydeus Hydeus is offline
Free Member
 
Join Date: Dec 2003
Location: Omicron Persei 8
Posts: 322
Thanks: 0
Thanked 0 Times in 0 Posts
I'm starting to be lost in posts

Two questions to Incredible:
1st few post ago by Dialhot
Quote:
Originally Posted by Dialhot
WHO actually set the matrix in bbmpeg ?
If I follow you correctly, if you do not do that, all the KVCD muxed with bbmpeg will be unscramble with the wrong key
2nd
Quote:
Originally Posted by Incredible
Min Quantizers I=3, P=3, B=4, 84 Frames within a GOP! CBR 1165, 1B Frame (yes it decreased filesize), noise added via ffvfw.
And what will be with no B frames at all?
__________________
Go for SECAM =)
Reply With Quote
  #316  
02-04-2004, 06:23 AM
Krassi Krassi is offline
Free Member
 
Join Date: Mar 2003
Location: Germany
Posts: 390
Thanks: 0
Thanked 0 Times in 0 Posts
I've done 3 screenshots.
Some details:
DVB-Capture mpeg2, small adapted MA-script, q=4 (all).
1 min. equals 18.283 KB.

Bitrate
Pic1
Pic2

In the last picture on the top right corner you can see some blocks. However, TMPGEnc's image is alot worse

EDIT: The screenshots above had SADT, adaptive 3. The following is done with "normal" settings:
Pic3

EDIT2: Converted image links to normal links to clean up thread.
Reply With Quote
  #317  
02-04-2004, 06:44 AM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
CAN ANYONE TELL ME WHY ME Q CURVE ISN'T CONSTANT

Krassi have a perfect line, Inc have a "near constant" curve and mine is like the yellow line of Krassi snapshot !

Whatever the settings I put (I even set to 100% all the paremeters in Rate control and put "0" for "Max Quantizer difference"

This sucks !

Note: Inc and all, still snapshots are worthless. They only permit to see if there is macro blocks default. But all other defaults have a temporal aspect (dancing block, ants, mosquitos).

What are we doing there ? Photo or video ? You can have perfect (or awfull) stills but awfull (or good) video.

The last sample of Kwag is a perfect example of what I want to say. I'm sorry to say that but it was very bad !
(see my post in the other thread talking about ffvfw).
Reply With Quote
  #318  
02-04-2004, 06:52 AM
Krassi Krassi is offline
Free Member
 
Join Date: Mar 2003
Location: Germany
Posts: 390
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Dialhot
Krassi have a perfect line, Inc have a "near constant" curve and mine is like the yellow line of Krassi snapshot !
Maybe the registry settings are damaged. Have you tried a reinstall? If i remember well you have already installed the latest build, right
Reply With Quote
  #319  
02-04-2004, 07:00 AM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Krassi
if i remember well you have already installed the latest build, right
Yes. And I will have to purge manually the register because I already uninstall/reinstall a lot with no real results. I was using ffdshow since months, that can be tied.

Note for Krassi: don't miss the ep. 10-11 tomorrow
Reply With Quote
  #320  
02-04-2004, 07:30 AM
rds_correia rds_correia is offline
Free Member
 
Join Date: Apr 2003
Location: Chinese Democracy starts now!
Posts: 2,563
Thanks: 1
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Dialhot
CAN ANYONE TELL ME WHY ME Q CURVE ISN'T CONSTANT

Krassi have a perfect line, Inc have a "near constant" curve and mine is like the yellow line of Krassi snapshot !
Hi guys,
Sorry to say but I have the same problem that Phil has about the Q curve not being constant.
Mine is looking like krassi's yellow one too...
Anyway I am having very good encoding results with ffvfw's settings posted on the other
thread, apart from this Q curve prob.
Has anyone noticed that speed decreased since we changed to constant bitrate?
Maybe not much on some PCs but I went from 17 hours encoding to 22 hours
encoding a 115 minutes movie.
Yep my pc is needing new brains
I'll be seeing u guys tonight.
__________________
Rui
Reply With Quote
Reply




Similar Threads
Thread Thread Starter Forum Replies Last Post
FFMPEG vs FFVFW vs Mencoder ? bilu Video Encoding and Conversion 561 04-15-2004 06:16 AM
FFMPEG: Observation about ffvfw poerschr Video Encoding and Conversion 28 02-24-2004 05:50 PM
FFMPEG: Ffvfw - What is video buffer verifier and what is it matter? Prodater64 Video Encoding and Conversion 1 02-23-2004 07:46 AM
FFMPEG: Do ffvfw and mencoder/ffmpeg give the same results? Razorblade2000 Video Encoding and Conversion 4 02-06-2004 04:23 PM
FFMPEG: XMPEG 5.03 and ffvfw kwag Video Encoding and Conversion 2 02-05-2004 10:57 AM

Thread Tools



 
All times are GMT -5. The time now is 05:46 AM  —  vBulletin © Jelsoft Enterprises Ltd