AviSynth TSMC (Temporal Soften Motion Compensated)(Denoiser)

[Jetrell Fo]

Okay, here's the deal, I want to denoise my SE LD Trilogy that I uploaded to Myspleen. Can I do this to them without changing anything of their audio/video specs? Do my videos have to be .avi for it to work properly?

[althor1138]

Yeah. I don't know what colorspace you are working in but mvtools2 requirements are listed on the wiki page. I know for a fact that it works with yv12 and yuy2. They don't have to be avi it was just an example.

[Jetrell Fo]

Would having VirtualDubMod on hand be an asset for this stuff as well?

[althor1138]

Virtualdub is almost a hindrance when it comes to avisynth stuff, or even editing imo. I just use it for capturing my laserdiscs to lags yuy2 codec. I usually do x264 encoding in megui or just open the avs file directly with x264 cli. I recommend installing avspmod and getting familiar with avisynth commands and filters. For me, it's capture in Virtualdub, edit in avspmod using avisynth.

[Chewtobacca]

Thanks for this denoiser, althor!  Do you have any thoughts on applying it to animation? I'd guess that a lower mthresh value (closer to 70 than 120) might be in order.

[althor1138]

I don't know much about denoising animation but if it's from an analog source I guess I would use a larger temporal radius (5 perhaps?) and then go as high as you can with mthresh without seeing unwanted artifacts and then dial it back 20 or 30. 70-120 was just an eyeball estimate of what would be safe thresholds that I decided upon for live action film. It could vary wildly from source to source. I used 180-230 when denoising the rotten Star Wars P&S laserdisc for example.

I have developed another version that allows you to manually control the blur on the prefiltered clip which helps prevent MVTOOLS from locking onto noise as "motion" which might help things further. I will update the first post with that when I get near my pc again this weekend.

[Chewtobacca]

Thanks!  Yeah, it's from an analog source.  I have already have the temporal radius on 5 because I'm following the first post.  So far, I've settled on the following.

Code:

TSMC(5,90,255,4,true)

It seems to do a good job without making the video look too scrubbed.  I wonder whether the fact that animation has so many duplicates will be a help or a hindrance. I don't have much experience with denoising...

[althor1138]

 I have already have the temporal radius on 5 because I'm following the first post.

I should change that to 3 as the default. While tsmc seems to usually do just fine with 5 as long as you've tweaked the mthresh to a level where it is picking up on coherent motion, 3 is probably safer and should denoise nearly as well.

 I wonder whether the fact that animation has so many duplicates will be a help or a hindrance. I don't have much experience with denoising...

I think that would help. If each frame has 1 duplicate, tsmc could be tweaked to only use the duplicates for denoising. Then you could crank up the mthresh as high as you want and simply blend the frames to remove noise. I will do a tsmc version aimed at anime denoising and see if it is viable or not.

[Chewtobacca]

I think that would help. If each frame has 1 duplicate, tsmc could be tweaked to only use the duplicates for denoising. Then you could crank up the mthresh as high as you want and simply blend the frames to remove noise. I will do a tsmc version aimed at anime denoising and see if it is viable or not.

Such a version would be a great asset. What might complicate matters is that occasionally animation has sections (usually those involving a high degree of motion) that are animated at 24fps, as opposed to the usual 12fps with every frame duplicated.

[spoRv]

Such a version would be a great asset.  What might complicate matters is that occasionally animation has sections (usually those involving a high degree of motion) that are animated at 24fps, as opposed to the usual 12fps with every frame duplicated.

About this, what's about the restoration idea to discard the duplicated frame, and insert interpolated ones?