Is there a way to sharpen video without sharpening the artifacts?

[Master Tape]

In Avisynth+? When i use Sharpen, it makes the picture look better and a bit more detailed, but the side effect is it highlights all compression artifacts, macroblocks and mosquito noise, ect. Is there any script that just focuses on sharpening the image and not all the other junk?

[butterw]

Re: Is there a way to sharpen video without sharpening the artifacts?
Not Really.
Other sharpening algorithms (there are plenty to choose from) may achieve better results on your source however.

An Edge sharpening method for instance would only attempt to sharpen detected edges.

[Master Tape]

I think i've found one that does the job. ASharp

Adaptive thresholding to avoid noise enhancement.
Block adaptive sharpening to avoid DCT block edges enhancement.
Block based adaptive thresholding to avoid DCT block edges enhancement.

But... i can't seem to get it to run with Avisynth+ looks like it was only designed for 32-bit versions of Avisynth.

[poisondeathray]

But... i can't seem to get it to run with Avisynth+ looks like it was only designed for 32-bit versions of Avisynth.

You can run almost any x86 avs filter in x64 avs+ with mp_pipeline

It's unlikely that you will avoid enhancing the noise with any single filter

[Master Tape]

But... i can't seem to get it to run with Avisynth+ looks like it was only designed for 32-bit versions of Avisynth.

You can run almost any x86 avs filter in x64 avs+ with mp_pipeline

It's unlikely that you will avoid enhancing the noise with any single filter

Can't get it to work. It says "MP_Pipeline: Can't find any splitter"

SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("QTGMC", 2)
MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_1.d2v", CPU2="ooooxx", Info=3)
AssumeTFF()
AssumeFPS(25)
QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true)
Spline64Resize(960,720)
Crop(0,0,-6,0)
AddBorders(0,0,6,0)
LoadPlugin("C:\Program Files\AviSynth+\plugins64+\MP_Pipeline.dll")
SetMemoryMax(1)
MP_Pipeline("""
### platform: win32
LoadPlugin("C:\Program Files (x86)\AviSynth\plugins\asharp.dll")
""")
Prefetch(5)

[poisondeathray]

You need to specify win64 and win32 sections , each divided up by ### ###

Because you're using CPU2= for mpeg2source, this requires the x86 version

The entire script that passes through win32/win64 sections has to be enclosed within MP_Pipeline


It would look like this

MP_Pipeline("""

### platform: win32
MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_1.d2v", CPU2="ooooxx", Info=3)
### ###

### platform: win64
AssumeTFF()
AssumeFPS(25)
QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true)
Spline64Resize(960,720)
Crop(0,0,-6,0)
AddBorders(0,0,6,0)
### ###

### platform: win32
LoadPlugin("C:\Program Files (x86)\AviSynth\plugins\asharp.dll")
asharp()
### ###
""")

The prefetch values probably need to be adjusted, because each process can use separate prefetch with thread chunking and branching. You can use avsmeter64 to play with it and see what works best. It's quite complex, the optimal value might not be simple as using one single prefetch value

[jagabo]

There are many ways to sharpen a video without sharpening "artifacts". What you do depends on the type and severity of artifacts. In some cases you can reduce the artifacts before sharpening (DCT ringing and blocking, grain, etc.). In others you can limit sharpening only to certain types of features using Overlay() and a mask. You need to provide a sample.

[Master Tape]

You need to specify win64 and win32 sections , each divided up by ### ###

Because you're using CPU2= for mpeg2source, this requires the x86 version

The entire script that passes through win32/win64 sections has to be enclosed within MP_Pipeline


It would look like this

MP_Pipeline("""

### platform: win32
MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_1.d2v", CPU2="ooooxx", Info=3)
### ###

### platform: win64
AssumeTFF()
AssumeFPS(25)
QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true)
Spline64Resize(960,720)
Crop(0,0,-6,0)
AddBorders(0,0,6,0)
### ###

### platform: win32
LoadPlugin("C:\Program Files (x86)\AviSynth\plugins\asharp.dll")
asharp()
### ###
""")

The prefetch values probably need to be adjusted, because each process can use separate prefetch with thread chunking and branching. You can use avsmeter64 to play with it and see what works best. It's quite complex, the optimal value might not be simple as using one single prefetch value

Thanks alot it's loading fine now. Though after having used it, it's highlighting the noise too much, noise that didn't look too noticeable before sharpening.

There are many ways to sharpen a video without sharpening "artifacts". What you do depends on the type and severity of artifacts. In some cases you can reduce the artifacts before sharpening (DCT ringing and blocking, grain, etc.). In others you can limit sharpening only to certain types of features using Overlay() and a mask. You need to provide a sample.

Yes this sounds like the best way to do it. Reduce the artifacts before applying Sharpen. But maybe a subtle less is more approach to improve the source without over-doing it.

I want to try using EzNoise with QTGMC, but i'm getting another script error "There is no function named 'sneo_FFT3D"

[poisondeathray]

I want to try using EzNoise with QTGMC, but i'm getting another script error "There is no function named 'sneo_FFT3D"

You need neoFFT3D and Zs_RF_Shared.avsi

https://forum.doom9.org/showthread.php?t=181294
https://forum.doom9.org/showthread.php?t=174121

[Master Tape]

I want to try using EzNoise with QTGMC, but i'm getting another script error "There is no function named 'sneo_FFT3D"

You need neoFFT3D and Zs_RF_Shared.avsi

https://forum.doom9.org/showthread.php?t=181294
https://forum.doom9.org/showthread.php?t=174121

Thanks poison. Still i'm not pleased with the result from EzDenoise, as Sharpening still highlights all the noise. However think i've finally found a sharpener that sharpens without sharpening all the noise aswell,

UnsharpMask(strength=32, radius=3, threshold=8)

Though i dialed down the strength a bit.

One more issue with the video i'm having is relating to the trims. There's a small segment in which a CGI source shimmers, and i apply TFM to it, but the trims just don't add up. I search through the video and make a note of where i want the trims.. add them then when i return(vid2) it goes to somewhere else in the video, like 4 minutes after from where i want it applied. And the shimmer remains as it hasn't been applied to the correct frames.

SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("QTGMC", 2)
MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_1.d2v", CPU2="ooooxx", Info=3)
AssumeTFF()
AssumeFPS(25)
vid1 = trim(0,12310).AssumeTFF().AssumeFPS(25).QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid2 = trim(12311,12577).AssumeTFF().AssumeFPS(25).TFM(pp=0).ChangeFPS(50).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid3 = trim(12147,32820).AssumeTFF().AssumeFPS(25).QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid1 ++ vid2 ++ vid3
Prefetch(5)

[poisondeathray]

One more issue with the video i'm having is relating to the trims. There's a small segment in which a CGI source shimmers, and i apply TFM to it, but the trims just don't add up. I search through the video and make a note of where i want the trims.. add them then when i return(vid2) it goes to somewhere else in the video, like 4 minutes after from where i want it applied. And the shimmer remains as it hasn't been applied to the correct frames.

SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("QTGMC", 2)
MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_1.d2v", CPU2="ooooxx", Info=3)
AssumeTFF()
AssumeFPS(25)
vid1 = trim(0,12310).AssumeTFF().AssumeFPS(25).QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid2 = trim(12311,12577).AssumeTFF().AssumeFPS(25).TFM(pp=0).ChangeFPS(50).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid3 = trim(12147,32820).AssumeTFF().AssumeFPS(25).QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid1 ++ vid2 ++ vid3
Prefetch(5)

The trims are applied to the original 25fps frame numbers. Not the 50fps frame numbers. Is that what you wanted ? If not, then adjust the trims

[Master Tape]

One more issue with the video i'm having is relating to the trims. There's a small segment in which a CGI source shimmers, and i apply TFM to it, but the trims just don't add up. I search through the video and make a note of where i want the trims.. add them then when i return(vid2) it goes to somewhere else in the video, like 4 minutes after from where i want it applied. And the shimmer remains as it hasn't been applied to the correct frames.

SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("QTGMC", 2)
MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_1.d2v", CPU2="ooooxx", Info=3)
AssumeTFF()
AssumeFPS(25)
vid1 = trim(0,12310).AssumeTFF().AssumeFPS(25).QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid2 = trim(12311,12577).AssumeTFF().AssumeFPS(25).TFM(pp=0).ChangeFPS(50).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid3 = trim(12147,32820).AssumeTFF().AssumeFPS(25).QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true).Crop(0,0,-6,-4).UnsharpMask(strength=32, radius=3, threshold=8).Spline64Resize(960,720)
vid1 ++ vid2 ++ vid3
Prefetch(5)

The trims are applied to the original 25fps frame numbers. Not the 50fps frame numbers. Is that what you wanted ? If not, then adjust the trims

Thanks, after much trial and error i got there in the end.

I've also since applied MosquitoNR and that seems to have helped a bit with the MPEG noise. But FFMpeg refuses to render with it in the line. Only when i remove it, it then renders fine.

[jagabo]

But FFMpeg refuses to render with it in the line. Only when i remove it, it then renders fine.

64bit vs 32bit? Multithreading issue?

Lots of others: http://avisynth.nl/index.php/External_filters#Deringing_.26_Mosquito_Noise

[Master Tape]

But FFMpeg refuses to render with it in the line. Only when i remove it, it then renders fine.

64bit vs 32bit? Multithreading issue?

Lots of others: http://avisynth.nl/index.php/External_filters#Deringing_.26_Mosquito_Noise

It seemed so, as using the 32-bit plugin and loading it via pipeline sorted it.

Also when you are loading from the index file, how do you load the audio from the file aswell? As don't want to remux it afterwards if i want to trim something frame accurately.

[poisondeathray]

The index for mpeg2source is only for video. Use an audio source filter with AudioDub(video, audio) . There are examples in the documentation

You can use other audio source filters such as FFAudioSource, or LWLibavAudioSource as well

If you are editing sections (e.g. cutting segments out, extending segments, etc..), you need to apply trims to audio as well, otherwise you will go out of sync

[Master Tape]

Hmm after i render, ffmpeg compresses the sound to only 128kbps, when the original audio is 384kbps. Is there a way to directly transcode the original audio but have AvsPmod trim it with the video?

[poisondeathray]

You can set the audio bitrate in ffmpeg e.g. -b:a 384k

If you use an avs script, it's (frame)serving uncompressed audio and video, so you are re-encoding both

If you're not doing edits that change the duration, you can keep the original audio (stream copy)

[Master Tape]

For some reason i'm getting a maximum of 360kbps, but it's not too much of a difference so can live with it.

I'll shift the -192ms sound delay in avidemux, since i think you can only offset the sound in ffmpeg when both video and audio are separate.

[poisondeathray]

What type of audio ? AAC ? Stereo ? 360kbps is a lot for stereo AAC

You can delay audio in the script too
http://avisynth.nl/index.php/DelayAudio

When you use DGIndex, usually the delay value is written into the name of the demuxed audio

You can preview the script for sync in mpchc, just comment out the slow filters temporarily

[Master Tape]

It is MP2 audio, from recordable DVDs.

Thanks, didn't know there was a script that could do that. Will come in useful.

Whilst everything is working fine normally, when i render with MP_Pipeline on, i'm getting no audio. And applying the trims to audio and video didn't help. I only need to apply the trims once as always, when not using MP_Pipeline.

MP_Pipeline("""

### platform: win64
SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("QTGMC", 2)
video = MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_2.d2v", CPU2="ooooxx", Info=3)
audio = FFAudioSource("F:\A Video Files\DVD Rips\VTS_01_2 Tc0 L2 2ch 48 384 DELAY -192ms.mp2")
AudioDub(video, audio)
DelayAudio(-0.192)

### ###
### platform: win32
LoadPlugin("C:\Program Files (x86)\AviSynth\plugins\MosquitoNR.dll")
MosquitoNR()
### ###
### platform: win64
AssumeTFF()
AssumeFPS(25)
QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true)
Crop(0,0,-6,-4)
UnsharpMask(strength=32, radius=3, threshold=8)
Spline64Resize(960,720)
Trim(15498,34887)
Prefetch(5)
### ###
""")

[poisondeathray]

Audio was not supported in the original mp_pipeline, it probably still isn't in the newer version

Bring the audio and audiodub outside of mp_pipeline

something like this

MP_Pipeline("""
.
.
.
""") 

vid = last

aud = FFAudioSource("F:\A Video Files\DVD Rips\VTS_01_2 Tc0 L2 2ch 48 384 DELAY -192ms.mp2")

AudioDub(video, audio)
DelayAudio(-0.192)

[Master Tape]

MP_Pipleline: Unable to create slave process. Message: Script error: Script error: Invalid arguements to function 'MPP_PrepareDownstreamClip'.

[poisondeathray]

MP_Pipeline("""

### platform: win64
SetFilterMTMode("DEFAULT_MT_MODE", 2)
SetFilterMTMode("QTGMC", 2)
MPEG2Source("F:\A Video Files\DVD Rips\VTS_01_2.d2v", CPU2="ooooxx", Info=3)

### ###
### platform: win32
LoadPlugin("C:\Program Files (x86)\AviSynth\plugins\MosquitoNR.dll")
MosquitoNR()
### ###
### platform: win64
AssumeTFF()
AssumeFPS(25)
QTGMC(preset="slow", matchpreset="slow", matchpreset2="slow", sourcematch=3, Lossless=2, TR2=2, EdiThreads=2, border=true)
Crop(0,0,-6,-4)
UnsharpMask(strength=32, radius=3, threshold=8)
Spline64Resize(960,720)
Trim(15498,34887)
Prefetch(5)
### ###
""") 

vid = last

aud = FFAudioSource("F:\A Video Files\DVD Rips\VTS_01_2 Tc0 L2 2ch 48 384 DELAY -192ms.mp2")

AudioDub(vid, aud)
DelayAudio(-0.192)

I don't know if that's ideal for multithreading. mp_pipeline has it's own branch threading and chunking. It takes a bit of testing to get the ideal values.

[jagabo]

Note that the 64 bit version of Mpeg2Source() doesn't implement the DCT deblocking and deringing filters. Try switching to the 32 bit version and you may not need MosquitoNR().

[Master Tape]

Note that the 64 bit version of Mpeg2Source() doesn't implement the DCT deblocking and deringing filters. Try switching to the 32 bit version and you may not need MosquitoNR().

Ohright i didn't realise. I can't preview it right now due to the error blocking it.


[Attachment 54710 - Click to enlarge]

Also it wasn't mosquito noise perse i was trying to eliminate, but dot crawl, since i didn't know what they were exactly, till i Googled.

[poisondeathray]

Set the MPEG2Source() section to 32bit to use the 32bit version

### platform: win32

[jagabo]

DCT deblocking won't help with dot crawl artifacts. Here's how you can use it with mp_pipeline though:

MP_Pipeline(""" 
### platform: win32 
Mpeg2Source("filename.d2v", CPU2="ooooxx", Info=3) 
### ### 
""") 

# 64 bit stuff

[Master Tape]

Solved the problem. I had to remove the "video =" reference next to MPEG2Source

[Master Tape]

How do you apply trims within MP_Pipeline? Seems like you have to do it a different way as just getting errors when i return(vid1) (vid2) ect. And with AudioDub() being outisde of it, makes things more confusing.

[jagabo]

Script?