Quantcast Avisynth: Smootherhiq vs. Convolution3D - Page 3 - digitalFAQ.com Forums [Archives]
  #41  
03-31-2003, 02:21 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
Hi KingTuk,

Try Convolution3D(preset="movieLQ") instead of STMedianFilter(8,15,4,7,8,15) in your script. Your file size will probably be just a little larger, but you'll preserve some fine details that are lost with STM. At least, that's what I was able to confirm, as viewed on the screenshots.

-kwag
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  
03-31-2003, 02:58 PM
KingTuk KingTuk is offline
Free Member
 
Join Date: Nov 2002
Posts: 107
Thanks: 0
Thanked 0 Times in 0 Posts
I tried that also and it did a good job and the file size was a little larger but it also took more time to encode...

have you tried a screenshot with these settings?

STMedianFilter(6,15,4,7,6,15)

I get the same file size as Convolution3D(preset="movieLQ") but in less time...
Reply With Quote
  #43  
03-31-2003, 06:15 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 KingTuk

have you tried a screenshot with these settings?

STMedianFilter(6,15,4,7,6,15)
No I haven't! I'll try it later

-kwag
Reply With Quote
  #44  
03-31-2003, 06:38 PM
Kane Kane is offline
Free Member
 
Join Date: Jan 2003
Posts: 398
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by KingTuk
DctFilter(1,1,1,1,1,.5,.5,0)
what is this filter suppose to make and where can i download it?
__________________
greetz Kane
Reply With Quote
  #45  
03-31-2003, 06:51 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
Hi KingTut,

Ok, the speed is definively faster: 5:10 for C3D, 4:28 for STM .
On my sample, file size is 86KB larger for the STM over C3D.
Quality is almost identical, with a slight advantage to STM over C3D, as small artifacts are less visible
And the high frequency details are maintained
So you have a winner: STMedianFilter(6,15,4,7,6,15)
Here's my script:

Code:
LegalClip()
GripCrop(528, 480, overscan=2,  source_anamorphic=false )
GripSize(resizer="BicubicResize")
mergechroma(blur(1.58))
mergeluma(blur(0.05))
SpaceDust()
TemporalSmoother(radius=2, strength=3)
TemporalCleaner(ythresh=5, cthresh=10)
STMedianFilter(6,15,4,7,6,15) 
DctFilter(1,1,1,1,1,.5,.5,0)
GripBorders()
LegalClip()
-kwag
Reply With Quote
  #46  
03-31-2003, 10:20 PM
dazedconfused dazedconfused is offline
Free Member
 
Join Date: Nov 2002
Posts: 316
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Kane
Quote:
Originally Posted by KingTuk
DctFilter(1,1,1,1,1,.5,.5,0)
what is this filter suppose to make and where can i download it?
It's supposed to "make" your .m1v smaller.

From the DCTFilter ReadMe:
Quote:
DctFilter is an experimental filter that, for each 8x8 block, will do a
Discrete Cosine Transform (DCT), scale down the selected frequency values, and then reverse the process with an Inverse Discrete Cosine Transform (IDCT).
Basically, this filter's probably not going to save you much space, but it's a very fast filter and really doesn't harm anything by using it, so you might as well. It can be downloaded here: http://www.jungleweb.net/~sansgrip/DctFilter_YUY2.zip . Disregard what it says in the ReadMe file about it only being an avisynth 2.5 filter (if this hasn't already been edited by now).

Quote:
Originally Posted by kwag
So you have a winner: STMedianFilter(6,15,4,7,6,15)
Looks like I've got some testing to do. Goodie goodie! Keep these script tweaks rolling in guys! At this point in the game, I think script-testing to find the premium compression-VS-quality tradeoff using the currently available filters is about all we really have left, isn't it? That is, if a "standard" all-around script is even a possiblity, what with the variance from source to source and all. It sure would be nice though.

-d&c
Reply With Quote
  #47  
04-01-2003, 03:22 AM
Kane Kane is offline
Free Member
 
Join Date: Jan 2003
Posts: 398
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by kwag
So you have a winner: STMedianFilter(6,15,4,7,6,15)
Looks like I've got some testing to do.

-d&c[/quote]

me, too!
thx for the info d&c
__________________
greetz Kane
Reply With Quote
  #48  
04-01-2003, 09:21 AM
Jellygoose Jellygoose is offline
Free Member
 
Join Date: Jun 2002
Location: Germany
Posts: 1,288
Thanks: 0
Thanked 0 Times in 0 Posts
Hi all!

With the current filters, I have gone back to good ol' 704x576 in resolution. with this resolution I can use sideborders of 24 pixel. I tried it on 2 TV sets, and I don't get black borders, so this is what I'm going to use! I think this resolution looks a lot sharper, and I still can up the CQ to around 70 for a 110 min. movie (Insomnia). In my eyes it looks better with this heavy filtering. just a little sharper than KVCDx3 resolution. you'll see the difference. here's the script i am using:
Code:
LoadPlugin("E:\MPEG-Tools\FitCD\MPEG2DEC.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\dustv5.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\legalclip.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\sampler.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\convolution3d.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\gripfit_preview.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\temporalcleanerold.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\dctfilter_YUY2.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\STMedianFilter.dll")

Mpeg2Source("C:\Insomnia\insomnia.d2v")

LegalClip()

GripCrop (704, 576, Overscan=0)

Gripsize()

TemporalSmoother(radius=2, strength=3) 
SpaceDust() 
TemporalCleaner(ythresh=5, cthresh=10) 
STMedianFilter(7,15,4,7,7,15)

mergechroma(blur(1.58))
mergeluma(blur(0.075))

GripBorders()
Letterbox(0,0,24,24)

DctFilter(1,1,1,1,1,.5,.5,0) 

LegalClip()

#Sampler(length=24)

ConvertToRGB24
With this resolution, you can go ahead and blur a little more, and maintain a sharp picture. however it's a matter of taste, I know. If anyone wants to try this, tell me your results.

btw: I think we've reached another small stair on the stairs to professional high-end encoding.
__________________
j3llyG0053
Reply With Quote
  #49  
04-01-2003, 10:05 AM
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 Jellygoose
btw: I think we've reached another small stair on the stairs to professional high-end encoding.
Yes, I think so too
btw, I see that you set overscan to 0 If you use 2 blocks, you'll be able to up the CQ even more

-kwag
Reply With Quote
  #50  
04-01-2003, 10:07 AM
Jellygoose Jellygoose is offline
Free Member
 
Join Date: Jun 2002
Location: Germany
Posts: 1,288
Thanks: 0
Thanked 0 Times in 0 Posts
My movie is widescreen. So I want to preserve maximum height. I even use 24 pixels overscan on each side, by just blacken them out with Letterbox. works fine for my tv at that resolution.
__________________
j3llyG0053
Reply With Quote
  #51  
04-01-2003, 10:17 AM
PyRoMaNiA PyRoMaNiA is offline
Free Member
 
Join Date: Jan 2003
Location: a PAL land (UK)
Posts: 408
Thanks: 0
Thanked 0 Times in 0 Posts
I like using 704x576 resolution, so I'll be using that script too!
Reply With Quote
  #52  
04-01-2003, 10:21 AM
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 Jellygoose
My movie is widescreen. So I want to preserve maximum height. I even use 24 pixels overscan on each side, by just blacken them out with Letterbox. works fine for my tv at that resolution.
Gotcha
And I think we can even push some ~120 minutes movies on one CD with these filter combinations

-kwag
Reply With Quote
  #53  
04-01-2003, 10:47 AM
KingTuk KingTuk is offline
Free Member
 
Join Date: Nov 2002
Posts: 107
Thanks: 0
Thanked 0 Times in 0 Posts
what benefit do we get by putting

mergechroma(blur(1.58 ))
mergeluma(blur(0.075 ))

at the end after the filters instead of before the filters and after the resize...

so why not this

LoadPlugin("E:\MPEG-Tools\FitCD\MPEG2DEC.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\dustv5.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\legalclip.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\sampler.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\convolution3d.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\gripfit_preview.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\temporalcleanerold.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\dctfilter_YUY2.dll")
LoadPlugin("E:\MPEG-Tools\FitCD\STMedianFilter.dll")
Mpeg2Source("C:\Insomnia\insomnia.d2v")
LegalClip()
GripCrop (704, 576, Overscan=0)
Gripsize()
mergechroma(blur(1.58 ))
mergeluma(blur(0.075 ))

TemporalSmoother(radius=2, strength=3)
SpaceDust()
TemporalCleaner(ythresh=5, cthresh=10)
STMedianFilter(7,15,4,7,7,15)
GripBorders()
Letterbox(0,0,24,24)
DctFilter(1,1,1,1,1,.5,.5,0)
LegalClip()
#Sampler(length=24)
ConvertToRGB24
Reply With Quote
  #54  
04-01-2003, 10:52 AM
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
Hi KingTuk,

The filter order you just posted is exactly what I'm using. It seems more logical (to me!) to do the color/brightness merge before applying noise, etc., filters, than applying after

-kwag
Reply With Quote
  #55  
04-01-2003, 01:26 PM
Jellygoose Jellygoose is offline
Free Member
 
Join Date: Jun 2002
Location: Germany
Posts: 1,288
Thanks: 0
Thanked 0 Times in 0 Posts
lol kwag. I just wanted to answer that this way seems more logical to me .
I thought it's more logical to smooth the picture first, and take the noise out of it, and after that blur the clean picture.
well i guess we have to wait for somebody special to answer this one (hint @ SansGrip) !
however I think it doesn't really change a whole lot. has anyone tried if the filesize changes?
__________________
j3llyG0053
Reply With Quote
  #56  
04-01-2003, 02:03 PM
KingTuk KingTuk is offline
Free Member
 
Join Date: Nov 2002
Posts: 107
Thanks: 0
Thanked 0 Times in 0 Posts
Well if you look at my one of my posts yesterday I used basically the same script and got a smaller file size with the mergechroma and mergeluma after the resize...

I don't know if by putting it at the end, after the filters if you get a better picture, but you do get a larger filesize...
Reply With Quote
  #57  
04-01-2003, 02:32 PM
ak47 ak47 is offline
Free Member
 
Join Date: Oct 2002
Posts: 168
Thanks: 0
Thanked 0 Times in 0 Posts
well this doesn't invole testing. But is the link for STMedianFilter link for the website is down or my computer sucks. If someone can give me the file I would be very thank full.
Reply With Quote
  #58  
04-01-2003, 03:00 PM
DorvalCS DorvalCS is offline
Free Member
 
Join Date: Dec 2002
Location: Montreal, Canada
Posts: 31
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via AIM to DorvalCS
Quote:
Originally Posted by kwag
Ok, the speed is definively faster: 5:10 for C3D, 4:28 for STM .
On my sample, file size is 86KB larger for the STM over C3D.
Quality is almost identical, with a slight advantage to STM over C3D, as small artifacts are less visible
And the high frequency details are maintained
So you have a winner: STMedianFilter(6,15,4,7,6,15)
Here's my script:

Code:
LegalClip()
GripCrop(528, 480, overscan=2,  source_anamorphic=false )
GripSize(resizer="BicubicResize")
mergechroma(blur(1.58))
mergeluma(blur(0.05))
SpaceDust()
TemporalSmoother(radius=2, strength=3)
TemporalCleaner(ythresh=5, cthresh=10)
STMedianFilter(6,15,4,7,6,15) 
DctFilter(1,1,1,1,1,.5,.5,0)
GripBorders()
LegalClip()
-kwag
Kwag,

I have a suggestion... Actually it's a request. I am amaze on how much these scripts change from one day to the other... And how much improvement we get from one version to the other. Seems to me this scripts are the key to good quality KVCD encoding. I try to come back to your site regularly to see if there are new and improved scripts that I can learn from and obvioulsy use... But the problem is I find it difficult to scan through all the threads to find out where you guys have come up with the "latest and greatest"...

Would it be possible for you to post the most current script somewhere on the site for us "lazy people" to go and retrieve? I think it merits a spot on your KVCD templates pages... I realize you can't possibly post your script everytime you make a change to it... Maybe once a week with the "Script of the week"???

That would be great....

You guys are amazing!!!

DorvalCS
Reply With Quote
  #59  
04-01-2003, 03:13 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 DorvalCS
Would it be possible for you to post the most current script somewhere on the site for us "lazy people" to go and retrieve?
Check the main forum page under "Optimal scripts"

-kwag
Reply With Quote
  #60  
04-01-2003, 04:06 PM
Jellygoose Jellygoose is offline
Free Member
 
Join Date: Jun 2002
Location: Germany
Posts: 1,288
Thanks: 0
Thanked 0 Times in 0 Posts
Great idea DorvalCS and kwag!
__________________
j3llyG0053
Reply With Quote
Reply




Similar Threads
Thread Thread Starter Forum Replies Last Post
Avisynth: Convolution3d on optimal script? dredj Avisynth Scripting 21 07-10-2003 03:13 AM
Avisynth: XP crashes on FaeryDust and Convolution3D? hachemoncourt Avisynth Scripting 8 01-30-2003 06:03 PM
Avisynth: Where to download vobsub.dll, convolution3d.dll, dustv4.dll nicksteel Avisynth Scripting 3 01-21-2003 01:52 PM
Avisynth: Check out the new Convolution3d LadyMiles Avisynth Scripting 8 11-03-2002 01:35 AM
Avisynth: BEST FILTER - temporal smoother or convolution3D? ? ARAGORN Avisynth Scripting 2 10-10-2002 04:38 PM

Thread Tools



 
All times are GMT -5. The time now is 04:52 AM  —  vBulletin © Jelsoft Enterprises Ltd