funky honeycomb interference problem

[R3Z]

Firstly just like to say hi and thanks to anyone reading this :)

I have been in the process of cleaning up a live show from a vhs transfer of a handycam. There are numerous issues such as dropouts, color bleeds, noise, residual combing and other glorious things - but the one i can seem to get rid of completely is the interference one.

Heres the source m2v (24mb);

http://www.users.on.net/~inet_s51/pref/VTS_01_2.demuxed.m2v

Heres an example;

http://www.users.on.net/~inet_s51/pref/int_n2.jpg
http://www.users.on.net/~inet_s51/pref/int_f2.jpg

Heres my script (for the interference);


DGDecode source ..

Bob(0,0.5).Deblock_QED()

f2quiver (
\"bp3",98,194,
\"bp3",96,-192,
\"bp3",96,64,
\"bp3",96,-64,
\"bp3",0,128,
\"bp3",195,0,
\"bp4",50,-96,
\"bp4",50,160)

Frequency spectrum from f2quiver before and after;

http://www.users.on.net/~inet_s51/pref/f2before.jpg

http://www.users.on.net/~inet_s51/pref/f2after.jpg

I have tried using defreq and f2quiver. No matter what i do, i cant get defreq to remove more than say 25% of the honeycomb effect. With f2quiver i can get say 80% done as seen by the screens. I have tried a few tricks to try and isolate the frequency hot spots like adjusting the gamma and using temporal soften after the f2quiver test function.

The honeycomb effect happens most in the cyan/blue colors.

Tried doing a mo comped medianblurt of the luma channel and removes this effect but also lots of detail.

Wondering if there is any way to use masktools to isolate the interference as a mask ? The interference seems to move very little.

Maybe what i have is the best i can get, but i thought it wouldnt hurt to seek help. I have much to learn and its great fun doing so.

Cheers for reading :D

[Leak]

Looks pretty much like dot crawl to me.

Search this forum for that term - there's a few quite sophisticated filters out there to combat this (DeDot comes to mind) that'll probably work even better than your solution.

np: Future Sound Of London - Long Shadows (From The Archives Vol. 3)

[R3Z]

Looks pretty much like dot crawl to me.

Search this forum for that term - there's a few quite sophisticated filters out there to combat this (DeDot comes to mind) that'll probably work even better than your solution.

np: Future Sound Of London - Long Shadows (From The Archives Vol. 3)

Thanks for your reply Leak :)

I thought it may be dotcrawl due to a PAL > NTSC conversion but i cant get dedot, guavacomb or Tcomb to even touch the sides. Tcomb gets rid of the dots, but in the process it makes it look like someone blured the picture and then oversharpened it.

If anyone else has any ideas i would be gratefull for your input :)