View Full Version : Any way to use extra field in IVTC for noise reduction?
Trixter
20th June 2023, 01:16
When processing telecined captures from NTSC VHS, the result of an ITVC is 4 unique frames, then a duplicate, etc; then a decimate process removes the duplicate, returning a 23.976 source.
This has always bothered me, as the extra field introduced in the telecine process could potentially be used to filter out the noise introduced by videotape. Has anyone explored this, either in a plugin or an avisynth process? I recall Donald Graft's Dup Filter, (https://www.rationalqm.us/dup/dupnew.html) which is appropriate for 12fps and lower anime, but was hoping for a way to do this for full 23.976 sources.
johnmeyer
20th June 2023, 18:23
Are you suggesting some sort of filter that would look at the two duplicates, decide which one was less noisy, and then decimate the less noisy of the two?
Having done a lot of IVTC over the years, for most situations, I don't think you'd notice any difference. However, I've seen a lot of videos posted over the years where one field was corrupt and an adjacent field was OK. Duplicating or estimating a field to fix the corrupt one can make a huge difference in that scenario.
Emulgator
20th June 2023, 23:44
The proposed suggestion will lead to unwanted results anyway:
Averaging from 2 undamaged frames will inadvertently generate the One frame which has less noise than the other three.
Noise Flicker. A second later you will search for a method to mend that...
lollo2
22nd June 2023, 11:04
With good captures of analog signal from NTSC VHS, in my case I experienced that there is no advantege in averaging noise in the way the OP propose.
As example, here a comparison of an analog capture, versus the result of the AviSynth Median filter (https://forum.videohelp.com/threads/362361-Median%28%29-plugin-for-Avisynth) processing, which takes 3 captures of the same segment and median them frame by frame. The final result shows no improvement.
https://www.youtube.com/watch?v=8NJrxD0I_tU&list=PLCvrZEXO1laG2zPYB6XHnA72SBFaL_yso&index=3
In addition, the "dancing" noise introduced by the VHS media, can be reduced in a very effective way using a good temporal filtering, i.e. TemporalDegrain2
Trixter
30th June 2023, 03:13
The proposed suggestion will lead to unwanted results anyway:
Averaging from 2 undamaged frames will inadvertently generate the One frame which has less noise than the other three.
Noise Flicker. A second later you will search for a method to mend that...
Okay, this is fair; anime would benefit, but with 24fps film material there could be "pulsing" of the noise, and that might be distracting. Well, I had to ask :-) Thanks for the replies.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.