toot: new "advanced-averaging" filter (ignores worst source)

[Reel.Deel]

I too would be very interested in a link (or links) to both TooT and ABdespot.

When I come across dead links I always check archive.org, half the time it's helpful. In this case, the source code for ABdespot seems to be archived (http://web.archive.org/web/20050929104845/http://e-rels.dyndns.org/downloads/) but not for Toot.

[Busty]

Another 4 years have passed and the archive doesn't host the codes for ABdespot nor toot.

I have these two Laserdiscs of the same concert, readily captured in lossless, and these plugins might be the ones to denoise these captures, but they are nowhere to be found.

So, here's another request to anyone who still might have them flying around somewhere. I know, chances are slim...

Thanks

[juhok]

VapourSynth implementation. Subject to change. https://forum.doom9.org/showthread.php?t=174723

I take this chance to address ChiDragon's post from 2013 and say that 2/3 is useful for VHS and other analog tape formats too in some cases.

[spoRv]

Busty, I have toot - it worked flawlessly on my previous PC with Windows XP, but not in the new Windows 10 - maybe it's just my fault... oh well, if you need it, drop me a PM.

[StainlessS]

SpoRv, can you post openly, I'm sure there are others that would like a copy and/or someone that wants to add to an archive. (esp if with source).

[althor1138]

Here's a 7 day link. I only have the dll.

https://we.tl/x7J4KRFisL

[StainlessS]

Thank you Althor1138, so much nicer that SpoRv. :)

[spoRv]

Didn't think for a moment that would be interesting for anybody but me and Busty... :D

Next time, I'll post a public link, promise!

[Busty]

Hey SpoRv,
thanks so much for this.

I realized later that for two identical LDs I'd need abdespot and not toot, but toot will come in handy for other captures I'll do.
Hopefully decomet can accomplish the same as abdespot.

[jmac698]

My decomet may actually help for this. It's designed to clean up flashing dots where it's different in two sources. (actually lines but dots works too).

Should be possible to do these in script now too.

I've also tried buying 3 copies of commercial tapes and did a median. It didn't work because one of the copies played really bad. You should be able to use this technique to get a source as good as the master though, with enough copies.

[Karyudo]

I can't believe it's been almost 19 years since I asked E-Male to code TooT for me! I still have a copy of TooT.dll (it's just 7kb), which anybody interested can grab from here:
https://drive.google.com/file/d/1L5k12ABJKD1gGmXst50ewORS8gloQ45q/view?usp=sharing

[Karyudo]

On my server I've also found ToTooT (Two or Three out of Three), which was an evolution of TooT that would average all three clips if the value of a pixel is within a certain user-specified difference in all clips.

The whole 165kb kit — with source and DLLs for both TooT and ToToot, and some sample AVS — is available here:
https://drive.google.com/file/d/1FYQNpJ6hmvGtth1A0hYz-kRKVdllDrJB/view?usp=sharing

[DTL]

Yes - AVS is about 20 years old.

And yes - the well known 'majority' idea in tech engineering still not lost and have even newer implementations. It is used also in aerospace engineering to auto-skip failed control data from several equal control paths before making 'final decision' in majority-selector. To increase reliability of mission-critical parts. If some control path is fail (less than half number of total control paths) - the correct decision still made from major-voting of all inputs on majory selector device. Also the system can mark the failed inputs for engineers to repair. Or log the error if failied state is auto-repaired.

Some years ago I suggest to local hardware manufacturers to create SDI-device for auto-fail-safe 'switching' for broadcast reliability: The 3-SDI input majority switcher analyse 3 time-synched input samples and select to output only major-looking. So if it is 3 SDI streams from 3 synched playback servers and 1 of 3 is died - the output stream keep running without any 1 sample damaged. But it was found that typically broadcast playback servers fail not too frequently (we have dead hang of SDI playback PC-type cards about several times per year resolved only by power-off reboot) and also it was found DV-based files from synched database of 3 servers or 3 playback ports not give 100% bit-equal stream. Sad poor design of software (may be DV recompression at copy or other issue). So this hardware idea still not implemented here.

Currently close idea is implemented in pmode=1 processing in MDegrainN - but it is now block-based and motion-compensated and not use averaging but only pure selection of 'best' looking block from any number of input blocks (so it is good to have >2, it is about auto-solved for any integer tr-value because number of blocks in tr-pool to analyse is 2* tr + 1 (2* tr and current frame). So it can create 100% temporal denoised output. Because any averaging attempt of 'better looking' selected blocks still got not 100% stabe output from small variations and rounding errors and so on.

The algorithm to select 'best' block from a set of input blocks currently used: Created a 2D table of dismetrics between each blocks pairs from input set (the diagonal of table is block with itself is 'skipped' also calculated only half of symmetry table and second half is copied) and calculated sum of rows (or may be columns - it should give the same result). And block indexed by the lowest sum is selected as 'best'. It can be also tried on per-sample processing in plugins like vsTTempSmooth (may be as additional processing mode).

For several different clips it can be tested with standard inverleave/process/selectevery scritping. So for 3 input clips the tr is 1.

Also motion-compensation may helps to decrease blurring and other effects in cases of jitter in several analogue captures of single media (tape of laserdisc).

Post an idea for new processing mode to vsTTempSmooth - https://github.com/Asd-g/AviSynth-vsTTempSmooth/issues/7