Looking for *Good* DeNoiser+DeBlocker combo

[Aquilonious]

I've been pulling my hair out lately on some movie backups I've been trying to clean up. With numerous searches on Doom9, I've tried more results than I can think of. (I have about 20 threads saved). When I remove grain I get blocking, when I remove blocking I get grain. :mad:

I have a collection of older DVD movies that I'm trying to clean up--2010: The Year We Make Contact, Logan's Run, and Rollerball are just a few examples.

What makes things very difficult is that I don't see the artifacts on my PC, but when I play the m2v(mpg) files on my Sony 32 NTSC WEGA TV, I can really point them out. I get blocking in dark areas

I have tried tried the following combinations and many, many more:

Undot+Deen [also tried Deen("a3d",4,9,1,9)

LoadPlugin("c:\other\plugins\Blockbuster.dll")
LoadPlugin("c:\other\plugins\DNR2.dll")
LoadPlugin("c:\other\plugins\Msharpen.dll")
Blockbuster (method = "noise")
Blockbuster (method = "blur")
DNR2()
Msharpen()

LoadPlugin("c:\other\plugins\Blockbuster.dll")
LoadPlugin("C:\Other\Plugins\LoadPluginEx.dll")
LoadPlugin("C:\Other\Plugins\DustV5.dll")
ConvertToYUY2()
PixieDust(5)
ConvertToYV12()
Blockbuster (method="noise")
Blockbuster (method="blur")

LoadPlugin("c:\other\plugins\Convolution3D.dll")
ConvertToYUY2()
Convolution3d (preset="movieLQ")
ConvertToYV12()

Cnr2()
eDeen(5,6,18,1,3,false) [also tried (7,14,21,2,3,false)]
converttoyuy2()
PeachSmoother() [also tried (NoiseReduction=50,Stability=0,DoSpatial=false)]
Converttoyv12()

I have used the above plugins as well as these with no luck:

DeNoise - Blocking artificats when removing grain
eDeen - little grain removal, blocking on higher settings
FFT - leaves grain and changes color hues
SpaceDust - very little grain removal
STMedian - too much loss of detail, even with Msharpen added
UnFilter - filters too much, either smoothing or graining with artifacts as a result
VagueDenoiser - similar to PixieDust--blocking [started with (threshold=0.8,method=1,nsteps=6,chroma= true) and worked my way up]

I've tried some others but have simply forgot and/or removed them. I'm using RB's Matrix Editor and Media Player Classic to experiment for results.

Is there a hosting site where I can post my results? I'd post mpg files rather than screen captures because you don't always see the artifacting in still frames.

[Pookie]

BlockBuster is ADDING noise to your target rather than removing blocks.


FunkyDeBlock - http://forum.doom9.org/showthread.php?t=72431&highlight=funkydeblock

Camembert - http://mf.creations.nl/avs/functions/?M=D
-----------------------------------------------------------------------
DegrainMedian by Fizick http://bag.hotmail.ru/

Removegrain by kassandro v1.0 http://home.arcor.de/kassandro/RemoveGrain/RemoveGrain.rar

[hartford]

I'm in NTSC land.

Cleaning depends on the source.

--
Old Black and White I use this often:

greyscale()

MVDenoise2()

--
But at other times I use this:

greyscale()

Pixiedust(limit=2)

LimitedSharpen()

--
On "3 Days of the Condor" I used:

Crop(0,58,-0,-62)

RemoveGrain(mode=1).RemoveGrain(mode=2)

ConvertToYUY2()
Pixiedust(limit=2)

ColorMatrix(mode="Rec.601->Rec.709")

ConvertToYV12()
LimitedSharpen(soft=true,smode=3)

AddBorders(0,58,0,62)

--

Pick your poison.

[hartford]

I've not had any "deblock" problems, but I have this script that claims to "fix" thins problem:

#Adjust transition_level to your likings

Loadplugin("d:\plugins\MaskTools-1.5.6.dll")
Loadplugin("d:\plugins\MPEG2Dec3dg.dll")

avisource("d:\Angel's_Share2.avi").converttoyv12()

normal = last

transition_level = 31

dark = BlindPP(cpu2="xxxxoo", quant=20, moderate_h=10, moderate_v=10)
#dark = dark.temporalsoften(2,7,11,15,2)
# dark = dark.greyscale() # perhaps even this

darkmask = normal.levels(16,1.0,transition_level,255,0,false).FitY2UV()

MaskedMerge(normal,dark,darkmask,Y=3,U=3,V=3)

#blindPP(quant=2,cpu=4)

Wish that I could provide more help.

[mg262]

I reckon you might be causing the problem by encoding at too low a bit rate?

Failing that, are you willing to wait? If so I would recommend trying TwiceDust (search for it)...

[Aquilonious]

I reckon you might be causing the problem by encoding at too low a bit rate?

Failing that, are you willing to wait? If so I would recommend trying TwiceDust (search for it)...
The source itself is blocky. PixieDust(2) leaves the best overall results but introduces blocky artifacting. Applying Blockbuster afterwards removes blocking but introduces the grain that PD had removed. My Bitrates average between 3500-4500.

I haven't tried TwiceDust or QuadDust. This is the only thread (other than this one) Doom9 returned on a search is:

http://forum.doom9.org/showthread.php?t=100592&highlight=TwiceDust

I tried Googling and came up with the same results.

[Leak]

Applying Blockbuster afterwards removes blocking but introduces the grain that PD had removed.
BlockBuster only tries to remove blocking produced by the *CODEC* you're encoding with by adding noise (or grain, as you call it) so the codec will use a lower quantizer to preserve the artificial "detail" that it thinks this noise is.

Try encoding with a lossless codec to really judge if it's your script that produces the blockiness or if it's the encoder; if it's the latter, up the bitrate, tweak your settings, use another codec or live with the blocks and deblock upon playback...

np: Stockfinster - Last Report (All Becomes Music)

[Aquilonious]

@hartford - I'm going to try what you mentioned, specifically the PD(2)+Lsharpen.

BlockBuster only tries to remove blocking produced by the *CODEC* you're encoding with by adding noise (or grain, as you call it) so the codec will use a lower quantizer to preserve the artificial "detail" that it thinks this noise is.

Try encoding with a lossless codec to really judge if it's your script that produces the blockiness or if it's the encoder; if it's the latter, up the bitrate, tweak your settings, use another codec or live with the blocks and deblock upon playback...

np: Stockfinster - Last Report (All Becomes Music)
I'm using only RB-Pro and CCE to backup original source files from the DVD (vob, ifo, bup). Unless RB-Pro uses some type of codec by default, there's no codec involved.

I'm still experimenting with Blockbuster. As I write this I'm using the following script on the movie 2010:

LoadPlugin("c:\other\plugins\Blockbuster.dll")
LoadPlugin("c:\other\plugins\LoadPluginEx.dll")
LoadPlugin("c:\other\plugins\DustV5.dll")
ConvertToYUY2().PixieDust().ConvertToYV12()
Blockbuster(method="noise",block_size=4,variance=0.3,seed=1)

So I'm still experimenting. There are several good suggestions in this thread that I'm going to try.

I'd like to try Didee's LTSMC script (http://www.neuron2.net/board/viewtopic.php?t=787), but I just can't get it to work. I don't know if I have to load something before the script or what not. Didee says the script's not ready for general use but from the results I've seen on really grainy source, it does a wonderful job.

I've attached some example bitmap files. The first is the source file, second is Deen, third is PD(2).

[Didée]

Are you getting an error message when loading a script with LTSMC? The script posted on that thread should contain everything that's needed.

For your topic here, also try this (http://forum.doom9.org/showthread.php?p=727667#post727667). Documentation is again sparse, but working it does quite good ;)

[mg262]

I'm using only RB-Pro and CCE to backup original source files from the DVD (vob, ifo, bup). Unless RB-Pro uses some type of codec by default, there's no codec involved.Well, there's the MPEG-2 produced by CCE. My Bitrates average between 3500-4500.could you just try doubling that and seeing how it looks?

QuadDust is in the thread Didée linked to... but try his method first. If you do try multi-dust, 4*dust is a pointless waste of time... 2*dust is just as good as the version I posted,* and probably as any 4*dust version.

*(I don't even remember being active on Doom9 that far back. I think that was my first real thread. Some of the stuff I said is really embarrassing!)

[Aquilonious]

Well, there's the MPEG-2 produced by CCE. could you just try doubling that and seeing how it looks?
Remeber, that's an average. I can't go higher without increasing the TargetSectors in RB and going to DVD-9.

I don't think simply doubling the bitrate is going to increase the quality of the source itself, which is on a DVD-5 pressed disc. You have to remember that the actual source is not all that great. Also, I've had some very clean encodes in the 3500-4500 bit range.

I guess my attachments didn't get posted. They were around 1.3mb bitmap files. I checked the forum rules and didn't see anything about a file size limitation. ???

@Didee - I was trying this: http://home.arcor.de/dhanselmann/_stuff/LTSMC.rar
I just saved the script you posted today. I'll try that once my current project is finished.

[mg262]

I wasn't suggesting using it for the real encode... just on a brief section to check that it wasn't responsible for (some of) the blocking. Whether you get blocking or not at ~4000 depends on the source -- it's perfectly possible that some sources will go through fine, and others (sharper or noisier or more grainy) will cause blocking. Since you can't see the effect on your PC, it's IMO worth checking this, if only to rule out the possibility.

[Aquilonious]

I wasn't suggesting using it for the real encode... just on a brief section to check that it wasn't responsible for (some of) the blocking. Whether you get blocking or not at ~4000 depends on the source -- it's perfectly possible that some sources will go through fine, and others (sharper or noisier or more grainy) will cause blocking. Since you can't see the effect on your PC, it's IMO worth checking this, if only to rule out the possibility.
Thanks for your quick feedback, mpg262. I know the source is grainy and blocky because I've played the original disc on several players with the same result. So without a doubt the source is low quality.

But you have a good idea here and one I will experiment with. I will temporarily change the Targetsectors in RB to 4472800 (see http://forum.doom9.org/showthread.php?p=724154#post724154) and using the FOX matrix with no filters, will see how a segment of the movie looks in RME.

By the way, I noticed here that you created a filter called BlockMedian. http://forum.doom9.org/showthread.php?p=724365#post724365
What kind of results have you got with it?

[Didée]

Just try BlockMedian(4), and get astounded :D

(It's meant for internal usage in more complicated scripts, not for direct image filtering.)

[Aquilonious]

Just try BlockMedian(4), and get astounded :D

(It's meant for internal usage in more complicated scripts, not for direct image filtering.)
Q. Should I run LTSMC after PD, or just by itself? The source has a LOT of grain and I want to remove the grain as well as blocking.

I've run the following tests:
1. Source
2. PD(2)
3. PD()
4. PD()+Blockbuster(method="noise",block_size=4,variance=0.3,seed=1)
5. LTSMC

Comments:
#2 - removes light grain but no blocking
#3 - removes heavier grain but very noticeable blocking
#4 - removes more grain & no blocking, perhaps the best (hard to say)
#5 - like #4, but produces just a slight green halo effect while transition from different shades of black. Big plus: Much faster than any of the other methods.

File Sizes: Source=77.8mb; PD2=67.8mb; PD5=66.3mb; PD5+Bb=86.3mb; LTSMC=86.3mb

Regarding LTSMC, the source also seems to exhibit the green halo effect but not as much.

It's like I'm scratching my head with a rake. :confused:

[Mug Funky]

hang on... your source is a DVD-5, so why not back it up by simply burning it onto a 5 as is? you can clean it later of course and re-burn if you have success with some filters.

i only recommend using blockbuster in extreme cases of blocking-in-darkness, and only with low settings. it doesn't conserve luma when it works on dark areas, and this can actually increase blocking (the blocks with noise added by blockbuster will be brighter than the blocks around them... this can be bad).

could you maybe upload a chunk of a particularly bad part of your source? there's enough scriptwizards around here to come up with something good.

also, you mention it looking fine on your PC but not on your TV. have you considered plugging the computer into the TV? it's extremely useful to know what things will look like on the teev before you waste hours encoding something. failing that, use this script (which i use when there's too much reflected light on the screen and dark areas are obscured):

function artefact (clip c)
{
c
separatefields()
converttorgb()
levels(0,1.6,255,0,255)
sharpen(1)
converttoyuy2()
weave()
}

just put this in your script (or somewhere in your plugins dir inside a .avsi file) and type "artefact()" to run it. this script is designed to brighten the picture and emulate a TV (but sharper).

this will highlight most common problems with blocking and ringing, and also helps with green/pink tints in dark areas.

DO NOT encode with this turned on - it's for spotting problems only, and WILL make your movie look worse than it is. comment it out with a # to get rid of it before you encode.

[mg262]

Mug Funky,

Wow that thing brings out the noise. It makes me think about the possibility of applying some transformation like that, then filtering, then applying the reverse transform ... I'm not sure whether this is a sensible idea or not. (All three steps in one plug-in running internally at high colour depth.)

[Aquilonious]

Mug Funky,

Wow that thing brings out the noise. It makes me think about the possibility of applying some transformation like that, then filtering, then applying the reverse transform ... I'm not sure whether this is a sensible idea or not. (All three steps in one plug-in running internally at high colour depth.)
I tried the artifact (correct spelling ;) ) script and agree with your comments here that it brings out a lot of noise. It also runs rather slow on my pc.

I have my old (circa '94) Sony 20" TV hooked up with its s-video input. The only problem is that the ATi drivers drive the TV at a 1024x768 resolution, so the TV overscans. It's the TV's max resolution. At that resolution I have a hard time discerning the artifacts. The only way I can run the TV at 800x600 is if I set my desktop to 800x600 as well, but ugggh--I don't want to do my work in an 800x600 res!

My monitor's a very good one--a Sony G420 19". I most always use a 1280x960 res which of course is also a 4:3 ratio (vs the odd 1280x1024). Even if I manipulate the brightness contrast I still can't see the blocking artifacts. Perhaps this is because it's running non-interlaced and the TV is running, of course, interlaced. I can experiment more but it's getting tedious to do all this.

I'm running the following script right now and it's taking a looong time (it's PD that's slowing it to a glacial speed).

LoadPlugin("c:\other\plugins\LoadPluginEx.dll")
LoadPlugin("c:\other\plugins\DustV5.dll")
LoadPlugin("c:\other\plugins\MaskTools.dll")
LoadPlugin("c:\other\plugins\RemoveGrainS.dll")
ConvertToYUY2().PixieDust(2).ConvertToYV12()
Import("c:\other\plugins\LTSMC.avs")

So far it's been running for over 15 hours. Currently the above script is 80% through a second pass running at 0.75 fps. The file size is about 450mb. Standalone, LTSMC runs much, much quicker--it never dips below 2.5 fps, even while encoding very large files, like the current one that's processing.

@Didee - Is your LTSMC script meant to run AFTER PixieDust, or should it be run standalone? If I should run it after PD, what limit should I set PD to? I use 2 but the default is 5. Although 5 produces more blocking it also reduces more grain. LTSMC is meant to reduce blocking after PD, right? I ask because I want to get a sense of how your script should be used.

Also, you say LTSMC's a wrapper for PD but you've lost me on this. I'm still relatively new at much of this and don't understand what you mean by "wrapper".

Could you or someone please explain? Thanks!

[Mug Funky]

http://dictionary.reference.com/search?q=artefact

:)

they're both correct, but i like to spell things the ye olde englishe waye (colour, centre, aluminium). it doesn't really matter of course... just good old pedantry from me.

if you want full size display on your TV, try using a hacked driver. i believe there's some for the ATI cards. you should be able to output 720x480 or 720x576 (the matrox cards can auto-detect and do both, which is totally sweet and i wish nVidia would do it). the TV encoder chip is outputting 480/576 lines anyway, so there's no reason the card can't actually do it.

that said, it can be a hellish experience getting a TV card to output standard TV...

the artefact filter is basically a substitute for a TV to give some idea of what would be visible on a good, well calibrated TV in a darkened room.

[Aquilonious]

http://dictionary.reference.com/search?q=artefact

:)

they're both correct, but i like to spell things the ye olde englishe waye (colour, centre, aluminium). it doesn't really matter of course... just good old pedantry from me.

if you want full size display on your TV, try using a hacked driver. i believe there's some for the ATI cards. you should be able to output 720x480 or 720x576 (the matrox cards can auto-detect and do both, which is totally sweet and i wish nVidia would do it). the TV encoder chip is outputting 480/576 lines anyway, so there's no reason the card can't actually do it.

that said, it can be a hellish experience getting a TV card to output standard TV...

the artefact filter is basically a substitute for a TV to give some idea of what would be visible on a good, well calibrated TV in a darkened room.
Thanks for the enlightenment on the word artefact. I use dictionary.com all the time and usually I look up a word to make sure, but I should know better. I email several Canadians who use colour, saviour, and other such spellings. I actually like the spellings over the Americanized versions because, pardon the pun, they have more colour. :D

That being said, I still don't know why Europeans use commas instead of decimal points. For example, 3,55 instead of 3.55. Perhaps you can shed some light on that. :p

The script in my previous post ended up taking 20 hrs and 3 min. Arrgh! I had some old National Georgraphic tape recordings I wanted to clean up and PD2+LTSMC really did a very good job. Some grain is still apparent, but it's the fault of the source material--I think you can only clean up an analog source so much before losing important detail.

As far as the ATi drivers are concerned, I've tried the "optimized" Omega drivers several times but always end up going back to the WHQL Catalyst drivers. I'm also a gamer and I just haven't had good luck with the Omega drivers (while others apparently have). I suppose it depends on what games/software one uses.

I'd desperately like to upgrade my system but just don't have the $$$ right now. I like the DFI Ultra-D SLI mobos but would proobably end up going for the DFI Infitiy SLI instead. Both ATi & nVidia have decent graphics cards. I've owned cards from both and I really have no preference, other than the fact that image quality is more important to me than the fastest frame rates.

I'd still like to hear from Didee on LTSMC so I'm going to send a pm to him.

[Didée]

Aquilonious, you're mixing up things :)

Is your LTSMC script meant to run AFTER PixieDust, or should it be run standalone?
...
Also, you say LTSMC's a wrapper for PD but you've lost me on this.
LTSMC is a standalone denoiser, and I would recommend to NOT use PixieDust when you're using LTSMC.

The "wrapper" for PixieDust is *not* LTSMC, but PixieDustPP :)
What I mean here by "wrapper" is that PixieDustPP is not a postprocessing filter to be run after PixieDust, but a complete solution doing [pre-processing + PixieDust + PostProcessing + AllNeededColorspaceConversions].

After some more testing (not very much though), I'd rather suggest using s=1 instead of s=3 in PixieDustPP.

[Aquilonious]

Aquilonious, you're mixing up things :)


LTSMC is a standalone denoiser, and I would recommend to NOT use PixieDust when you're using LTSMC.

The "wrapper" for PixieDust is *not* LTSMC, but PixieDustPP :)
What I mean here by "wrapper" is that PixieDustPP is not a postprocessing filter to be run after PixieDust, but a complete solution doing [pre-processing + PixieDust + PostProcessing + AllNeededColorspaceConversions].

After some more testing (not very much though), I'd rather suggest using s=1 instead of s=3 in PixieDustPP.
I think I'm going to try a faster degrainer than PD. It simply takes too long on my system to do tests with it, even with small clips.

A few Q's:

1. What do you think of the combination of DeGrainMedian+LimitedSharpen as opposed to PixieDustPP? Is PD better than DGM for heavy grain removal?

I've tried DGM on several sources and so far as I can tell, it produces slightly "softer" results compared to PD(5) but without blocking. I've tried both the 2-stage and

2. What DLLs do I need to load to run LimitedSharpen?

3. What is the basic difference between LimitedSharpen and Msharpen, other than the latter being slower?

[Poutnik]

when I remove blocking I get grain. :mad:
I have not tried it myself yet, processing mostly analog TV captures,
but it could be worthy to give a try to Kassandros's MDeblock.

http://www.mdeblock.de.tf/

Discussion about his filters are are invited in his forum

http://videoprocessing.11.forumer.com/

[Didée]

A few Q's:

1. What do you think of the combination of DeGrainMedian+LimitedSharpen as opposed to PixieDustPP? Is PD better than DGM for heavy grain removal? PixieDust is technically more complicated and complex, hence slower but in theory "better". Blocking is an issue of controlling the complicated-ness and complexibility ;)
Choose your poison.

2. What DLLs do I need to load to run LimitedSharpen?
MaskTools (http://manao4.free.fr/MaskTools-v1.5.8.zip) , WarpSharpPackage (http://www.avisynth.org/warpenterprises/files/warpsharppackage_25_dll_20031103.zip) for Smode=1.

3. What is the basic difference between LimitedSharpen and Msharpen, other than the latter being slower? Basically, MSharpen uses edge masking to apply [normal] sharpening only to edges.
LimitedSharpen uses a form of sharpening where the sharpening effect gets limited to not create edge halos, and to not emphasize noise. Edge Masking similar to MSharpen can be used too, as can "inverse" edgemasking.

Let's take the samples on neuron2's MSharpen page (http://www.neuron2.net/msharpen/msharpen.html) as an example, and compare closely to LimitedSharpen:

Source -- MSharpen(128) -- LimitedSharpen(strength=175,overshoot=0,undershoot=16)

http://img456.imageshack.us/img456/1082/hinasource29kq.th.jpg (http://img456.imageshack.us/my.php?image=hinasource29kq.jpg) http://img456.imageshack.us/img456/6236/hinamsharpen1282is.th.jpg (http://img456.imageshack.us/my.php?image=hinamsharpen1282is.jpg) http://img475.imageshack.us/img475/9453/hinalimitedsharpen9vh.th.jpg (http://img475.imageshack.us/my.php?image=hinalimitedsharpen9vh.jpg)
(All three re-saved from Vdub)

[danpos]

@Didée

I do see clearly that using LimitedSharpen you got rid of 'haloes' around edges from boy's white 'skirt' (?), which doesn't happen with MSharpen use.

Really a good comparison. :)

Regards,

EDIT: Didée, which LTS version did you use because the one that I use haven't the "undershoot" parameter ?

[Aquilonious]

PixieDust is technically more complicated and complex, hence slower but in theory "better". Blocking is an issue of controlling the complicated-ness and complexibility ;)
Choose your poison.


MaskTools (http://manao4.free.fr/MaskTools-v1.5.8.zip) , WarpSharpPackage (http://www.avisynth.org/warpenterprises/files/warpsharppackage_25_dll_20031103.zip) for Smode=1.

Basically, MSharpen uses edge masking to apply [normal] sharpening only to edges.
LimitedSharpen uses a form of sharpening where the sharpening effect gets limited to not create edge halos, and to not emphasize noise. Edge Masking similar to MSharpen can be used too, as can "inverse" edgemasking.


Thanks for posting the pictures, Didee. Though I don't do much with animation, the pictures describe what you can't in words. I've had similar effects with Msharpen so I use it sparingly. I prefer Dust to DGM, but Dust runs so slow. I'm still in search of a combo of filters to use on grainy material that will run "reasonably" well on my old (at least it seems that way!) Athlon 3200.

I have posted a new thread because I'm trying to get advice on how to clean up some grainy movies I have. I've included pictures of the material that I'd like to filter. If you can spare a moment please offer any suggestions you may have. Thanks a bunch. :)

[Fizick]

For comparison see also fft3dfilter(bt=-1,sharpen=1.0) picture:
http://www.avisynth.org/fizick/fft3dfilter/hinafft3dfilter.jpg

Aquiloninous,
you say:
FFT - leaves grain and changes color hues
How it changes the color hue?

[Revgen]

Aquiloninous,
you say:
FFT - leaves grain and changes color hues
How it changes the color hue?

He may be talking about this. (http://forum.doom9.org/showthread.php?p=595080#post595080)

But that was a long time ago. I haven't had any hue issues with the latest versions.

[Aquilonious]

For comparison see also fft3dfilter(bt=-1,sharpen=1.0) picture:
http://www.avisynth.org/fizick/fft3dfilter/hinafft3dfilter.jpg

Aquiloninous,
you say:
FFT - leaves grain and changes color hues
How it changes the color hue?
Revgen is right. I was going by an older thread in regards to the color comment. My sincere apologies--I should have noted the date of the posting! With the latest version of FFT3D I do not see any color issues. So I stand corrected. :)

FFT3D, like Dust, runs very slow on my system (I'm using the latest version, 1.6). Even with the following I'm only getting about 0.41 fps on my first CCE pass. I use 2 passes.

FFT3DFilter(sigma=3, interlaced=true)

I'm looking for a filter for light-moderate grain removal that's a bit faster than Dust but equal in quality. The nearest I've found is DeGrainMedian, which boasts excellent speed but sometimes too much loss of detail. I'll keep experimenting.

[Fizick]

0.41 fps?
You must have al least 7 fps with one FFT3Dfilter and your system config!
Or you use large (HD) frame size?

And latest fft3filter is version 1.8.1 now :)

[Boulder]

I think he means 0.41 * real time (~10fps) as that's what CCE shows. That's pretty much IMHO, my interlaced encodes with LTSMC run ~0.1RT on my 3.22GHz P4 :p

[acrespo]

0.41 fps?
You must have al least 7 fps with one FFT3Dfilter and your system config!
Or you use large (HD) frame size?

And latest fft3filter is version 1.8.1 now :)

I have a similar config and FFT3DFilter is 3 fps only. The fft3dfilter line in my scripts is:
fft3dfilter(sigma=8, plane=0, sharpen=0.7)
A anime with 25 minutes

My config:
ASUS A7N8X-X (nForce2)
Athlon XP 3200+
1GB RAM; 2.5-3-3-5
Geforce 6600GT 128MB
Maxtor MaxLine III 250GB SATA
Soundblaster Audigy 2 ZS
WinXP Pro SP2
Latest BIOS, driver, & WinXP critical updates installed
AviSynth 2.5.6

[Didée]

As far as the thread title's "Deblocking" is concerned, perhaps this (http://videoprocessing.11.forumer.com/viewtopic.php?p=413#413) is something, too ;)

[Boulder]

I have a similar config and FFT3DFilter is 3 fps only. The fft3dfilter line in my scripts is:
fft3dfilter(sigma=8, plane=0, sharpen=0.7)
A anime with 25 minutes

My config:
ASUS A7N8X-X (nForce2)
Athlon XP 3200+
1GB RAM; 2.5-3-3-5
Geforce 6600GT 128MB
Maxtor MaxLine III 250GB SATA
Soundblaster Audigy 2 ZS
WinXP Pro SP2
Latest BIOS, driver, & WinXP critical updates installed
AviSynth 2.5.6
What's the rest of your script?

[acrespo]

What's the rest of your script?
I don't have my script here but it's something like:

AviSource()
TFM(PP=7)
Tdecimate(mode=1)

Crop(align=true) #I dont remember the numbers
FFT3DFilter(sigma=8,plane=0, sharpen=0.7)
SpLine16Resize(640,480)
Dup() #I dont rembember the parameters too :P

[Aquilonious]

0.41 fps?
You must have al least 7 fps with one FFT3Dfilter and your system config!
Or you use large (HD) frame size?

And latest fft3filter is version 1.8.1 now :)
@Fizick -
I got v1.6 from here:

http://www.avisynth.org/warpenterprises/

I went to your web site and downloaded v1.8.1. Thanks for the pointing that out to me. :)

[Boulder]

I don't have my script here but it's something like:

AviSource()
TFM(PP=7)
Tdecimate(mode=1)

Crop(align=true) #I dont remember the numbers
FFT3DFilter(sigma=8,plane=0, sharpen=0.7)
SpLine16Resize(640,480)
Dup() #I dont rembember the parameters too :P
TFM and TDecimate will eat CPU cycles as well as resizing and Dup so I don't think FFT3DFilter is that slow ;)

[Poutnik]

....... I got v1.6 from here: http://www.avisynth.org/warpenterprises/ .....
Be careful with the site. In spite of its very high value due to a big list, they are often older versions. If you download plugin from there, check for newer version if possible.