Quantcast Avisynth: is there a Filter for Increasing Compression Even More? - Page 3 - digitalFAQ.com Forums [Archives]
  #41  
07-25-2003, 12:49 AM
Boulder Boulder is offline
Free Member
 
Join Date: Sep 2002
Location: Lahti, Finland
Posts: 1,652
Thanks: 0
Thanked 0 Times in 0 Posts
IMO DCTFilter should be the last item in the script, even after the borders. The author suggested this when he released the filter. I haven't done any filesize compares but I trust him
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
  #42  
07-25-2003, 06:43 AM
Holomatrix Holomatrix is offline
Free Member
 
Join Date: Aug 2002
Posts: 130
Thanks: 0
Thanked 0 Times in 0 Posts
Thanks all I'll try the script right away. I see your not using Limiter() and scd_trigger What is the reason for that?
Reply With Quote
  #43  
07-26-2003, 02:40 PM
Holomatrix Holomatrix is offline
Free Member
 
Join Date: Aug 2002
Posts: 130
Thanks: 0
Thanked 0 Times in 0 Posts
Ok, guys, check this script out By using Deen() + DCTFilter(1,1,1,1,1,1,0.5,0) file size goes way down while video quality goes up

nf=0
GripCrop(480,480,overscan=2,source_anamorphic=fals e)
GripSize(resizer="BiCubicResize")
Undot()
Deen()
Asharp(1,4)
STMedianFilter(8,32,0,0)
MergeChroma(blur(1.52))
MergeLuma(blur(0.2))
SwitchThreshold=(Width<=352)?4Width<=480)?3:2
ScriptClip("nf=round(YDifferenceToNext())"+chr(13) +"nf>=SwitchThreshold?unfilter(-(fmin(nf*2,100)),-(fmin(nf*2,100))):TemporalCleaner(6+nf,11+nf)")
GripBorders()
function fmin(int f1,int f2){return(f1<f2)?f1:f2}
DCTFilter(1,1,1,1,1,1,0.5,0)

Let me know how you like it
Reply With Quote
  #44  
07-26-2003, 03:47 PM
jorel jorel is offline
Invalid Email / Banned / Spammer
 
Join Date: Aug 2002
Location: Brasil - MG - third stone from the sun
Posts: 5,570
Thanks: 0
Thanked 0 Times in 0 Posts
oh yes Holomatrix,

i did tests with deen and it's very cool.

but it don't make your encode too slow?
Reply With Quote
  #45  
07-26-2003, 05:20 PM
Holomatrix Holomatrix is offline
Free Member
 
Join Date: Aug 2002
Posts: 130
Thanks: 0
Thanked 0 Times in 0 Posts
Found no speed decrease If any it would from DCTfilter but very small.
Reply With Quote
  #46  
07-26-2003, 06:07 PM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
Deen is a convolution filter as is StMedianFilter. Using both is perhaps "too much". For sure file size goes down, but I fear details level also.
Reply With Quote
  #47  
07-27-2003, 01:53 PM
Jellygoose Jellygoose is offline
Free Member
 
Join Date: Jun 2002
Location: Germany
Posts: 1,288
Thanks: 0
Thanked 0 Times in 0 Posts
You're also using a fairly high Luma-Blur Value in your script (0.2), which will for sure lead to a much smaller filesize, however, as Phil said, details will not be preserved very well the way I see it.
__________________
j3llyG0053
Reply With Quote
  #48  
07-27-2003, 07:35 PM
Holomatrix Holomatrix is offline
Free Member
 
Join Date: Aug 2002
Posts: 130
Thanks: 0
Thanked 0 Times in 0 Posts
I find that I still get tiny blocks in the video when I look close, is there anything in the script that I can modify to blend these blocks better? Blockbuster(dither) does not seem to help enough. Since I don't really know what all the commands in the script mean can I add scd_trigger or Limiter or up some values in the script? Thanks
Reply With Quote
  #49  
08-02-2003, 11:36 PM
J-Wo J-Wo is offline
Free Member
 
Join Date: Nov 2002
Location: Toronto, Canada
Posts: 454
Thanks: 0
Thanked 0 Times in 0 Posts
Question for both Jorel and Holomatrix. I noticed that your Undot and Asharp lines were both after Gripfit, but I thought Jorel had already found that they are better placed *before* resizing (like in Kwag's MA script). Any reason why they're back down there now?
Reply With Quote
  #50  
08-02-2003, 11:52 PM
jorel jorel is offline
Invalid Email / Banned / Spammer
 
Join Date: Aug 2002
Location: Brasil - MG - third stone from the sun
Posts: 5,570
Thanks: 0
Thanked 0 Times in 0 Posts
J-Wo my friend,
in the script that i posted i don't change the place of the filters.
and i'm using that script...asharp and undot are before Gripfit
giving better quality in the image and less size!

Reply With Quote
  #51  
08-03-2003, 12:03 AM
J-Wo J-Wo is offline
Free Member
 
Join Date: Nov 2002
Location: Toronto, Canada
Posts: 454
Thanks: 0
Thanked 0 Times in 0 Posts
Okay Jorel, now I'm confused! In page 1 of this thread you posted the following code:
Code:
GripCrop(480,480,overscan=1,source_anamorphic=false,dest_anamorphic=false)
GripSize(resizer="BiCubicResize")
Undot()
Asharp(1,4)
So here you have Undot and Asharp AFTER Gripcrop and Gripsize. Is this an error then?
Reply With Quote
  #52  
08-03-2003, 12:27 AM
jorel jorel is offline
Invalid Email / Banned / Spammer
 
Join Date: Aug 2002
Location: Brasil - MG - third stone from the sun
Posts: 5,570
Thanks: 0
Thanked 0 Times in 0 Posts
oh yes, you're not confused and you're right !
d&c(gentle giant) where are you?
change your username with me d&c!

i forgot,sorry J-Wo!
i update the script like Kwag's MA but using temporalcleaner,
like Phil recomendations,then....
take undot and asharp to before Gripcrop.....like this:

nf=0
Undot()
Asharp(1,4)
GripCrop(480,480,overscan=1,source_anamorphic=fals e,dest_anamorphic=false)
GripSize(resizer="BiCubicResize")
.....etc!

this is right!
Reply With Quote
  #53  
08-03-2003, 07:05 AM
Holomatrix Holomatrix is offline
Free Member
 
Join Date: Aug 2002
Posts: 130
Thanks: 0
Thanked 0 Times in 0 Posts
I didn't think it really made a big difference but to each his own. What was the reason we took out Limiter() ?
Reply With Quote
  #54  
08-03-2003, 07:59 AM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
Because I decide to and Jorel stuck to my advice .

I don't remember where I justified that choice but the idea is that TMPGEnc already does this limitation by defautl. So, why doing this twice ?

If you want to know where it's done in tmpgenc, it's there : settings, quantize matrix, output YUV as Basic YCbCr and not CCIR601. Unchecked (the defautl) = same effect than Limiter().
Reply With Quote
  #55  
08-03-2003, 08:27 AM
Holomatrix Holomatrix is offline
Free Member
 
Join Date: Aug 2002
Posts: 130
Thanks: 0
Thanked 0 Times in 0 Posts
Ok, so I guess I'll leave it out then and keep that option unchecked. KWAG still has it in his optimal script that's why I was wondering.
Thanks
Reply With Quote
  #56  
08-03-2003, 08:57 AM
jorel jorel is offline
Invalid Email / Banned / Spammer
 
Join Date: Aug 2002
Location: Brasil - MG - third stone from the sun
Posts: 5,570
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Dialhot
Because I decide to and Jorel stuck to my advice .

I don't remember where I justified that choice but the idea is that TMPGEnc already does this limitation by defautl. So, why doing this twice ?

If you want to know where it's done in tmpgenc, it's there : settings, quantize matrix, output YUV as Basic YCbCr and not CCIR601. Unchecked (the defautl) = same effect than Limiter().
Holomatrix,

the Phil explanations about limiter are here in this same thread,page 2:

http://www.kvcd.net/forum/viewtopic....r=asc&start=16

Reply With Quote
  #57  
08-03-2003, 10:24 AM
Wilbert Wilbert is offline
Invalid Email / Banned / Spammer
 
Join Date: Jun 2003
Posts: 45
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
the Phil explanations about limiter are here in this same thread,page 2:
I don't see what explanation you are referring to but "Limiter" is entirely different as "output YUV as Basic YCbCr and not CCIR601". The first one just round [0,16] to 16 and [236,255] to 236, while the latter _scales_ [16,236] to [0,255] when leaving it checked (and not scaled when leaving it unchecked).
Reply With Quote
  #58  
08-03-2003, 12:41 PM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
I don't know where you find that TMPGEnc scales up and not just clip values (if you can tell me where it's writen ). Btw, beside this thin diff, and despite what you seem to tell (do I misunderstand it ?), the result is the same : you end with values ranged from [16,235]. It's clearly said in the balloon that pops up on the screen when you let the cursor on the checkbox for a while (actually, with tmpgenc the lower limit is 8 and not 16).

So the question is still the same : why doing this twice ?
Reply With Quote
  #59  
08-03-2003, 01:15 PM
Wilbert Wilbert is offline
Invalid Email / Banned / Spammer
 
Join Date: Jun 2003
Posts: 45
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Btw, beside this thin diff, and despite what you seem to tell (do I misunderstand it ?), the result is the same : you end with values ranged from [16,235].
The resulting range is the same, but the result itself not (the values are distributed differently).

example: suppose pixel 1 has luma 24. With limiter this value remains 24. With the TMPGEnc checkbox it is changed to

y = x/2 + 8 [just an approx., don't know the exact formula]

x = 24 gives y = 24/2 + 8 = 20

Quote:
I don't know where you find that TMPGEnc scales up and not just clip values (if you can tell me where it's writen
I don't have the manual. But you always can do some tests to check this.
Reply With Quote
  #60  
08-03-2003, 02:37 PM
Dialhot Dialhot is offline
Free Member
 
Join Date: May 2003
Posts: 10,463
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
I don't know where you find that TMPGEnc scales up and not just clip values (if you can tell me where it's writen
I don't have the manual. But you always can do some tests to check this.[/quote]

So it's just a supposition ? But it's not that is said in the pop-up help (balloon). The balloon said exactly this : "Then Y Mpeg range is 8-235 (and the help talk about 16-235 !), not 0-255". I don't see there any evidence of rescaling. So ? Did you make yourself any check or do you just take this out of your brain ?

BTW the correct formula is :

if input is [0,255] and output is [8,235], the right formula is :

y=x/255 * ( 235-8 ) + 8 = 0.89 x + 8

So x=24 becomes y= 29. What was I called a thin diff, but you're right, it is not so thin.

The problem there is to know who is right.
Reply With Quote
Reply




Similar Threads
Thread Thread Starter Forum Replies Last Post
Avisynth: New compression filters? tengo6dedos Avisynth Scripting 17 04-27-2008 02:56 PM
Increasing brightness using avisynth? khusru Avisynth Scripting 7 02-23-2004 07:44 AM
Avisynth: Good compression? muhali3 Avisynth Scripting 10 10-14-2003 05:18 PM
Avisynth: MergeChroma to gain Compression? Jellygoose Avisynth Scripting 94 03-24-2003 02:27 AM
Avisynth: Increasing compression through preprocessing SansGrip Avisynth Scripting 67 01-29-2003 11:00 AM

Thread Tools



 
All times are GMT -5. The time now is 08:21 AM  —  vBulletin © Jelsoft Enterprises Ltd