digitalFAQ.com Forums [Archives] - Avisynth: Wavelet Noise Reduction?

Page 1 of 5

Show 40 post(s) from this thread on one page

digitalFAQ.com Forums [Archives] (http://www.digitalfaq.com/archives/)

- Avisynth Scripting (http://www.digitalfaq.com/archives/avisynth/)

- - Avisynth: Wavelet Noise Reduction? (http://www.digitalfaq.com/archives/avisynth/4317-avisynth-wavelet-noise.html)

kwag	07-03-2003 08:14 PM

Avisynth: Wavelet Noise Reduction?

I've been playing since yesterday with a filter that's floating around the net, which there is no English documentation. Here's the filter: www.kvcd.net/wnr.zip

It's a VirtualDub filter, but it can be made to work with AviSynth. The quality is just amazing, but it's SLOW and I mean SLOOOOWWWW.
But here it is so you can try it, and see for yourself the quality gained and file size dropped:

Just add these lines to your script:

Code:

# Include the following line where you normally load your plugins.

LoadVirtualdubPlugin("c:\wnr.vdf", "VD_wnr", 0)

#

Mpeg2Source("Your_Source_Here")

ConvertToRGB()

VD_wnr(7710, 7454, 7710, 0, 0, 0, 0, 0, 0)

ConvertToYV12()

#

# The rest of your script.

undot() 

Limiter()

Etc., ,etc., etc.

Here are instructions on how to get parameters from Vdub, and change them to use with AviSynth scripts: http://www.avisynth.org/index.php?pa...rom+VirtualDub

-kwag

kwag	07-03-2003 08:34 PM

Let's start the fun :mrgreen:

This will apply wavelet noise reduction ONLY where temporalsoften is being applied.

Code:

ScriptClip("nf = YDifferenceToNext()"+chr(13)+ "nf >= 2 ? \ 

unfilter( -(fmin(round(nf)*2, 100)), -(fmin(round(nf)*2, 100)) ) : \ 

TemporalSoften(4, round(1/nf) , round(3/nf) ,0, 2).ConvertToRGB().VD_wnr(7710, 7454, 7710, 0, 0, 0, 0, 0, 0).ConvertToYV12() ")

Now the speed increase is GiGANTIC :!:
The values 7710, 7454, 7710 are experimental :!:

Have fun ;)
-kwag

kwag	07-03-2003 08:41 PM

Corrected to this:

ConvertToRGB().VD_wnr( 20, 20, 20, 0, 0, 0, 0, 0, 0)

Code:

ScriptClip("nf = YDifferenceToNext()"+chr(13)+ "nf >= 2 ? \ 

unfilter( -(fmin(round(nf)*2, 100)), -(fmin(round(nf)*2, 100)) ) : \ 

TemporalSoften(4, round(1/nf) , round(3/nf) ,0, 2).ConvertToRGB().VD_wnr( 20, 20, 20, 0, 0, 0, 0, 0, 0).ConvertToYV12() ")

-kwag

ovg64

07-03-2003 09:40 PM

Re: Wavelet Noise Reduction

Quote:

English Documents inside zip here: http://www.avisynth.org/~warpenterprises/

kwag	07-03-2003 10:56 PM

This filter doesn't work correctly in a dynamic adaptive nature :!:
It was fun to test, but the current adaptive script filter combination just beats the crap out of it :mrgreen:
Bye Bye WNR :!:

Edit: Using it alone as a static filter, it does work, but after comparing clips with the current script, the file size is way smaller with the current script WITHOUT the WNR filter.

-kwag

sh0dan

07-04-2003 04:42 AM

I have no idea if this is anything even slightly similar:

VagueDenoiser 0.11: Wavelet based denoiser.

But you might give it a whirl.

kwag	07-04-2003 10:03 AM

Thanks sh0dan,

Apparently it is :!:
I just can't get the thing to work.
It calls for two parameters, but I get "Invalid argument to function "VagueDenoiser" when I try to open it in Vdub.
Time to look at the source 8)

-kwag

ovg64

07-04-2003 10:12 AM

Yeah I cann't get to work either :!: :?

kwag	07-04-2003 11:34 AM

Ok, I got it working. I had the .dll in the plugins directory of AviSynth, and I guess autoload didn't work correctly.
I specified the hard path, and now it works.

LoadPlugin("C:\Filters25\VagueDenoiser.dll") as usual, and then VagueDenoiser(0.8, 2) after the source line ( that's where I'm testing ).

-kwag

kwag	07-04-2003 12:07 PM

Well, here are the results. The filter is SUPERB :!: Look at the file size drop on the 9 second clip 8O
But now we need a 10Ghz computer to encode :mrgreen:

Here are the benchmarks and samples:

Current adaptive script ONLY:
http://www.kvcd.net/nonr.mpg
Encode time: 23s. Clip size: 9 seconds. File size: 1,083KB

Current adaptive script+wavelet noise filter VagueDenoiser(0.8, 1):
http://www.kvcd.net/nr1.mpg
Encode time: 1m:59s. Clip size: 9 seconds. File size: 1,053KB

Current adaptive script+wavelet noise filter VagueDenoiser(0.8, 2):
http://www.kvcd.net/nr2.mpg
Encode time: 1m:44s. Clip size: 9 seconds. File size: 1,055KB

-kwag

Jellygoose

07-04-2003 12:33 PM

Quote:

Originally Posted by kwag

Time to get a new CPU... 8)
I'd love to try this out on a full movie since I'm gonna be out of town until tomorrow evening. :D

kwag	07-04-2003 12:37 PM

Quote:

Originally Posted by Jellygoose

Time to get a new CPU... 8)
Where in the current script would you put that line? I'd love to try this out on a full movie since I'm gonna be out of town until tomorrow evening. :D

:oops: Sorry I forgot to mention that :!:
Right after your source line, before resizing. You can try it after resizing too. I haven't yet :!:

Edit: Works like a charm AFTER GripSize :!: Time dropped to 1m:5s :mrgreen:

-kwag

kwag	07-04-2003 12:45 PM

Anyone who wants to try the script, here it is:

Code:

## DLL Section ## 

# 

LoadPlugin("C:\Filters25\MPEG2Dec3.dll") 

LoadPlugin("C:\Filters25\GripFit_YV12.dll") 

LoadPlugin("C:\Filters25\STMedianFilter.dll") 

LoadPlugin("C:\Filters25\asharp.dll") 

LoadPlugin("C:\Filters25\unfilter.dll") 

LoadPlugin("C:\Filters25\undot.dll") 

LoadPlugin("C:\Filters25\VagueDenoiser.dll") 

# 

#### 



## Defined Variables and Constants ## 

# 

MaxTreshold = 1.50 

nf =  0 # Current frame. 

# 

#### 



## Main section and static filters ### 

# 

Mpeg2Source("Your_D2V_Source_Here") 

# 

undot() 

Limiter() 

asharp(2, 4) 

GripCrop(Your_GripCrop_Parameters_Here) 

GripSize(resizer="BicubicResize") 

VagueDenoiser(0.8,2)

STMedianFilter(8, 32, 0, 0 ) 

MergeChroma(blur(MaxTreshold)) 

MergeLuma(blur(0.2)) 

# 

# 



## Linear Motion Adaptive Filtering ## 

# 

# ( Portions from AviSynth's manual ) - This will apply temporalsoften to 

# very static scenes, and apply variable blur on moving scenes. 

# We also assign a variable - and this is why a line break is inserted: 



ScriptClip("nf = YDifferenceToNext()"+chr(13)+ "nf >= 2 ? \ 

unfilter( -(fmin(round(nf)*2, 100)), -(fmin(round(nf)*2, 100)) ) : \ 

TemporalSoften( fmin( round(2/nf), 6), round(1/nf) , round(3/nf) ,0, 2) ") 



# 

# 

# 



GripBorders() 

#LetterBox( Your_Values_Here ) # Depends on situation. Use MovieStacker! 

Limiter() 



# 

# 

## Functions ### 



function fmin( int f1, int f2) { 

  return ( f1<f2 ) ? f1 : f2 

} 



# 

####

-kwag

kwag	07-04-2003 01:26 PM

And for using the wavelet filter only on the adaptive part of the script, use this:

Code:

ScriptClip("nf = YDifferenceToNext()"+chr(13)+ "nf >= 2 ? \ 

unfilter( -(fmin(round(nf)*2, 100)), -(fmin(round(nf)*2, 100)) ) : \ 

TemporalSoften( fmin( round(2/nf), 6), round(1/nf) , round(3/nf) ,0, 2).VagueDenoiser(0.8,2) ")

That will speed things up quite a bit ;)
Remember to remove the VagueDenoiser(0.8,2) on the top of your script after GripSize ;)

Edit: The last parameter of VagueDenoiser (Type of wavelet) must be 0 or 1.

From the docs:

"Two types of wavelets are used here:
-Symmlet8: Daniel Lemire algorithm (use wavelet=0)
-Daubechies4: Ian Kaplan algorithm (use wavelet=1)"

So use 0 or 1. Not 2 as I had tested :!:

-kwag

kwag	07-04-2003 02:08 PM

Here's a sample with adaptive wavelet filter in full action.
You'll see the value of "nf" on the top of the screen, and remember that when the value is greater or equal to 2, the wavelet and temporal filters are not operational. Only on values below 2 is that the temporal and the wavelet filters are functioning. I used wavelet type 1 (Ian Kaplan algorithm), because the sample size was 1KB smaller that with type 0 (Daniel Lemire algorithm)
http://www.kvcd.net/adaptive-vague-1.mpg

Happy testing ;),
-kwag

vhelp

07-04-2003 02:24 PM

@ Kwag..

sheesh.. gimmie a chance to D/L your first sample :)

Edit: just got fist clip. Looks great here.

Now on to others..

I'll give your latest scrip-snip a go shortly.

Thanks for the newbie items.
-vhelp

vhelp

07-04-2003 02:28 PM

Kwag..

What class does that Filter fall under ??

Color space:
* YUV, YV12

AVIsynth:
* v2.52, v2.50 or older v2x

So that I can plut in its right place, where I keep my DLL's ie,
* \dlls\.. for old v2x, and
* \dlls252\.. for latest YV12 and AVIsynth v2.52

Tanks,
-vhelp

kwag	07-04-2003 02:30 PM

From the docs:

"Required:
Avisynth 2.52
YV12 colorspace"

-kwag

phibertron

07-04-2003 04:57 PM

Quote:

Originally Posted by kwag

Code:

ScriptClip("nf = YDifferenceToNext()"+chr(13)+ "nf >= 2 ? \ 

unfilter( -(fmin(round(nf)*2, 100)), -(fmin(round(nf)*2, 100)) ) : \ 

TemporalSoften( fmin( round(2/nf), 6), round(1/nf) , round(3/nf) ,0, 2).VagueDenoiser(0.8,2) ")

i downloaded the filter, but it is up .13 now
and tried the above, VagueDenoiser(0.8,1)
either .13 has a memory leak, or gremlins are hard at work

my systems page file kept growing and growing and growing...
it got up to like 4Gigs and my system ran out of memory

if i dont put it in the adaptive section, then my system only uses 250Mb of page file, and it works fine
so maybe it isnt the gremlins

just figured i would pass this along

-phibertron

vhelp

07-04-2003 05:07 PM

Yes, I agree.

Definate "memory leak". I got to the point where my system crashed in
one instance, and while another, I got a warning that my HD was filled and
to close apps and things. naturally, I stop TMPG's encoding. This was only
based on a 1 minute clip I was test encoding (which took approx. 5 minutes
to complete)

But I am noticing something in the final quality sage. "chizels" on solid
scenes where you can see "hatches" or "blocks" float around. I'm still
debuggin this, as it's still to "bright" in my workroom - - its ben sunny
all day here. Its harder to judge w/ too much sun is shining hehe..

-vhelp

Page 1 of 5

Show 40 post(s) from this thread on one page