02-03-2004, 03:23 PM
|
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
NICE! By this I can now directly test the optimal qantisizing for encoding! Extreme coool!
|
And now you should see clearly a MUCH better quality on the walls
The encoder was not optimized, because it was encoding with an erroneous VBV buffer size. So now look at the DCT blocks on the walls, and I'm sure they are equal (if not better?) that CCE
-kwag
|
Someday, 12:01 PM
|
|
Site Staff / Ad Manager
|
|
Join Date: Dec 2002
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
|
|
|
02-03-2004, 03:27 PM
|
Free Member
|
|
Join Date: Nov 2002
Posts: 863
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Quote:
Originally Posted by incredible
Did you also check in the left right corner of your task bar, that little FFvfw tab!! There you can see the graph while encoding and see how much the quantisation will take!!!
|
Where exactly on which screen is this?
And Kwag, you made my day with the screen examples!
|
02-03-2004, 03:27 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Ok, Kwag but your way in the last pics with that field where you put 10 and when changing than would give filesize differencies, .. won't work for me.
But what works is when changing the Qauntisation values from 2 to 3 in the I P and B frame quantisation section ... and that makes sense?!
|
02-03-2004, 03:31 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
|
02-03-2004, 03:42 PM
|
Free Member
|
|
Join Date: Apr 2003
Location: Chinese Democracy starts now!
Posts: 2,563
Thanks: 1
Thanked 0 Times in 0 Posts
|
|
@Nicksteel,
You can find it on the "statistics" settings of ffvfw.
Just choose show tray/graph.
Hurry man. Go run some tests with it
C ya
__________________
Rui
|
02-03-2004, 04:01 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
323 kb image size jpeg quality 50% ----- who cares, lets kick da bandwith!
Ok, here are the samples at PAL 25 FPS 704x576 overscan=2
1% Sample Filesize: 8.360 kb! Watch the avg bitrate in Bitrateviewer at 704x576/25fps! And thats a 1.85:1 movie.
Movie length about 95 mins, let it be more who cares if it wont get over 1300mb á 3 Movies on one DVD-R non-anamorphic.
Here it comes, first Bitrateviewer followed by the jpeg Pic (Quality 50%!!)
NEXT Step will be full CVD compilant 352x576 mpeg2 one movie to one CD using Notch! incl. 128kbit audio ... lets see!
|
02-03-2004, 04:09 PM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
SWEEEEEEET Incredible
I'm with you 100%: Bye Bye file size prediction
-kwag
|
02-03-2004, 04:17 PM
|
Free Member
|
|
Join Date: Mar 2003
Location: Germany
Posts: 390
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Great Result, Inc.
With or without filters
I have to find a DVD thats clean enough for this comparison. Xmen II is too noisy
My experiences with Karl's latest settings are similar.
The bitrate peaks go up to 6500 while average is ~2000. This is just great
|
02-03-2004, 04:23 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Here's the script I used:
Code:
mpeg2source("G:\One_HOUR_Photo\photo.d2v")
Asharp(1,4)
BicubicResize(656,368,0,0.6,0,4,720,536)
STmedianFilter(3,3,0,0).TemporalSoften(2,5,8,15,2).undot()
MaJustSoften()
AddBorders(24,96,24,112)
DctFilterD(4)
DctFilter(1,1,1,1,1,1,0.5,0)
Slicer(1,15,2,2,0)
I used Also DCTfilterD(4) as the last values of the MA in the right korner below of the matrix won't get higher than 63! But on the other hand no merge/chroma/blur
|
02-03-2004, 04:28 PM
|
Free Member
|
|
Join Date: Mar 2003
Location: Germany
Posts: 390
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
@inc: Thanks for the info.
@Kwag:
Changing the min/max quant. doesn't change filesize or average bitrate for me. I'll do some further tests, but there wasn't any difference between 2, 3 and 10.
|
02-03-2004, 04:34 PM
|
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 Krassi
@Kwag:
Changing the min/max quant. doesn't change filesize or average bitrate for me. I'll do some further tests, but there wasn't any difference between 2, 3 and 10.
|
It has to
For example, start with all values set to 2.
If file is too large, change them ALL to 3. Etc, etc.
The higher the numbers, the lower the file size.
Works for me
-kwag
|
02-03-2004, 04:40 PM
|
Free Member
|
|
Join Date: Mar 2003
Location: Germany
Posts: 390
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Quote:
Originally Posted by kwag
If file is too large, change them ALL to 3. Etc, etc.
The higher the numbers, the lower the file size.
Works for me
|
Yep, you're right. I missed "All"
I thought that min would be a minimum and max a maximum. So you have to change both. Do you know why this makes sence
|
02-03-2004, 04:43 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
@ Krassi
Quantisation means in ffvfw "Q" mode, means the quality, the more you quantize the more the pic gets worse.
The max quantize Values of I P and B is just ignored by teh encoder, min Qantize is what counts, do look during encoding at the graph as I explained above and you can see that the quantizer while encoding stays almost constant at the min value you did set in the quantize options! And will be the same then in Bitrateviewer!
Now at 352x576 the encoder behaves a bit different!! and not that efficient Like at bigger image sizes!
I did lower the min quantizers back to the better value of 2! (I P B) but on the other hand I can go to 2510kbit CBR (= VBV 44) (lower values will let suffer the image! and we should keep in mind that 352x576 will be scaled 2xtimes in its with afterwards!)
Heres the report on 352x576 at still 15 GOP!: Final size on 1% sample is 6.164 incl. Audio would fit nice on one CDr80. Next step will be 480x576!
|
02-03-2004, 04:48 PM
|
Free Member
|
|
Join Date: Mar 2003
Location: Germany
Posts: 390
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Quote:
Originally Posted by incredible
@ Krassi
Quantisation means in ffvfw "Q" mode, means the quality, the more you quantize the more the pic gets worse.
The max quantize Values of I P and B is just ignored by teh encoder, min Qantize is what counts, do look during encoding at the graph as I explained above and you can see that the quantizer while encoding stays almost constant at the min value you did set in the quantize options! And will be the same then in Bitrateviewer!
|
Okay, got it now. Thanks.
|
02-03-2004, 04:54 PM
|
Free Member
|
|
Join Date: Jul 2003
Location: Valencia (España)
Posts: 741
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
@incredible:
will we need a different CBR kbps value for each resolution?. Do you advise 2510 for 352x576 res.
|
02-03-2004, 04:56 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
In my workout now its the mixture of min Qantize and the kbit value in CBR!
But till now I do not recommend nothing as these are my first successes and Im doing on 480x576 now.
|
02-03-2004, 04:57 PM
|
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 digitall.doc
@incredible:
will we need a different CBR kbps value for each resolution?. Do you advise 2510 for 352x576 res.
|
I'll answer that.
*** NO ***
The CBR bitrate value controls the size of the VBV buffer size.
Leave the values that I posted on the procedure to make KVCDs or KDVDs.
Edit: Correction. You can actually go up to 48KB ( a VBV value of 24 for KVCDs and VCDs ) which should be optimal.
-kwag
|
02-03-2004, 04:59 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Kwag! In my case it controls the VBV Buffer size AND the value of avg bitrate! And in my samples it makes sense as only adjusting the min quantizer wont give effective results as the baviour of resulting filesizes got a too wide range in just one step of quantize value
|
02-03-2004, 05:01 PM
|
Free Member
|
|
Join Date: Apr 2002
Location: Puerto Rico, USA
Posts: 13,537
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Cross posting/editing
|
02-03-2004, 05:04 PM
|
Free Member
|
|
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
|
|
Strange! When doing 352x576 I could rise a bit the endsize/Quality by screwing at the CBR but now at 480x576 the final size is FIX!
|
All times are GMT -5. The time now is 06:52 PM — vBulletin © Jelsoft Enterprises Ltd
|