Non-Removeable Jaggies

[Alwyn]

Here we go again. I've digitised a VHS tape, which appears to be PSF.

There are major jaggies on the coloured wing of the aeroplane at the 1.3 second mark. Using Direct Stream Copy, I've added a pan of another aeroplane alert which doesn't seem to be as bad.

I used TFM on it, to no avail, and then QTGMC and VDub's internal deinterlacer.

Is there a solution or have the jaggies been burnt in by the maker of the tape?

Thanks for looking!

[lollo]

Some times ago you were asking for an example why analysis of fields is more appropriate than frames for checking video architecture. Here you have one. The jaggies are already in the fields, they do not depend on the fields architecture.

QTGMC for progessive repair does not help much here, but IIRC if you search the forum jagabo had sourced in the past some script to reduce such a defect

[poisondeathray]

The general approach is to apply antialiasing filter(s) on that section only (e.g. use Trim) if the entire video does not need it - because AA filters are quite destructive to detail in general

Since they jaggies are quite severe, one trick is to downscale, apply AA filter(s), then upscale

If you have other sections, you can just apply the filtered version with more Trims, or RemapFrames

o=AVISource("Aero Jaggies snip.avi").ConvertToYV16() #for some reason clip is RGB

o
Spline16Resize(400,300)
santiagmod(3,3)
QTGMC(inputtype=2)
Spline64Resize(720,576)
f=last

o.trim(0,33) + f.trim(33,209) + o.trim(210,0)
#ConvertToYV12()  #for the demo

[Sharc]

@Alwyn: Any reason why you captured to RGB?

[lollo]

Since they jaggies are quite severe, one trick is to downscale, apply AA filter(s), then upscale

Excellent solution, master pdr (similar to jagabo IIFC)!

Any reason why you captured to RGB

Yes, I also noticed that and had to convertToYUY2() for processing, strange

[Alwyn]

Thanks PDR, I couldn't get SantiagMod working (got it from here) but Santiag looks great.

Re SantiagMod, I got a "nnedi2 -32 bit filter" error message. I'm running 64bit and there doesn't appear to be a 64bit nnedi2.

I'll now do a CBA on whether it's worth going through the whole video to fix it. There are bits that are progressive or PSF so it will be a bit of work.

I'll have a look at Trim.

Some times ago you were asking for an example why analysis of fields is more appropriate than frames for checking video architecture. Here you have one. The jaggies are already in the fields, they do not depend on the fields architecture.

No I wasn't. I was querying why I should use Separatefields verses a quicker Double-rate deinterlace with VDub to analyse the fields.

I hadn't actually understood/seen this hard-jaggy situation in the fields before now so it has been a learning exercise.

Any reason why you captured to RGB?

Blame HUFF. I forgot I had dramas with this before. The only reason I use HUFF is to keep you guys happy. Never again! I'll use UT video next time.

[poisondeathray]

Thanks PDR, I couldn't get SantiagMod working (got it from here) but Santiag looks great.

Re SantiagMod, I got a "nnedi2 -32 bit filter" error message. I'm running 64bit and there doesn't appear to be a 64bit nnedi2.

Any version should be ok and give similar results. The version I was using is dogways'

https://github.com/Dogway/Avisynth-Scripts/blob/master/EX%20mods/SantiagMod.avsi

[Alwyn]

Yes, that one worked. Needed the "CL" and the bin as well. Thanks!

[lollo]

Some times ago you were asking for an example why analysis of fields is more appropriate than frames for checking video architecture. Here you have one. The jaggies are already in the fields, they do not depend on the fields architecture.

No I wasn't. I was querying why I should use Separatefields verses a quicker Double-rate deinterlace with VDub to analyse the fields.

Which is "checking video architecture".

In this example if you double deinterlace you can get confused by the defect (somehow similar to the interlaced effect) and apply, as you did, TFM, QTGMC, VDub's internal deinterlacer, etc.

If you analyze the fields rather than the deinterlaced frames you see immediately the problem.

[Sharc]

Any reason why you captured to RGB?

Blame HUFF. I forgot I had dramas with this before. The only reason I use HUFF is to keep you guys happy. Never again! I'll use UT video next time.

Maybe check your HUFFY configuration ("we guys" are happy with any other lossless, no need to worry)

[Attachment 82657 - Click to enlarge]

[lollo]

This is mine (I am happy as well):