Combining 2 analog captures to produce a better result
 

I have just seen this https://www.vegascreativesoftware.in...recipe--20943/
He takes 2 analog captures of the same video and combines them by effectively averaging to produce a better result.
Has anyone done this with Virtualdub or AviSynth? - is it possible with these tools?
What were the results like?

Avisynth has been doing that for years.


Median Avisynth plugin and filter package, explanation, syntax, functions:
http://avisynth.nl/index.php/Median



image examples:
https://forum.doom9.org/showpost.php...5&postcount=16

original 2014 doom9 thread:
https://forum.doom9.org/showthread.php?t=170216

This worked well for me: https://forum.fanres.com/thread-1299.html

I don't know how much respect it pays to the fidelity of the original source since I believe it invokes some interpolation on top of the averaging -- though I guess the averaging is already fudging things to begin with -- but it resulting levels were generally more consistent and less noisy.

The time consuming part is doctoring the various sources so that all of the frames match up. Chains of DuplicateFrame and DeleteFrame.

On those examples - it does a great job on the white streaks, but looking at the median example, the averaging of the middle two values (from four captures) doesn't seem to do much else. The noise looks a tiny bit better but not enough for me to want to do multiple captures.
Has anyone seen any substantial difference when there are no gross aberrations like those white streaks?

The images are early examples. They were made almost 5 years ago. The median filter has been improved since then, and several plugins have been derived that use similar frame averaging techniques using only one video (one is called FixRipsP2, which has been referenced here in several threads. An example is the baby video sample posted yesterday: http://www.digitalfaq.com/forum/vide...html#post58792. ) Those images were followed in that doom9 thread by another 4 years of discussions and revisions.



The method isn't that easy and has varying results with different videos and different noise patterns. Rest assured, if it was that easy and worked all the time in all situations for all users, everybody would be doing it. They aren't.