Algorithm for headswitching noise detection

Algorithm for headswitching noise detection

Postby silicontrip » 30 May 2013, 04:33

Hi Jan,

You refer to an old feature of your software as the "bottom garbage" feature.

I'm working on a project for detecting (and possibly removing) common artefacts in video digitisation.
Head switching noise is one of them. http://preservation.bavc.org/artifactat ... hing_Noise

Did you have any references on how you detected this noise? Can I ask how your algorithm worked?

Thanks
Mark
silicontrip
New
 
Posts: 1
Joined: 29 May 2013, 02:58
Location: Melbourne

Re: Algorithm for headswitching noise detection

Postby jeschot » 30 May 2013, 19:52

Hello Mark,

There is still the v1.2.4 version on my site which can fix the bottom garbage.
This was only for the JVC 9600. The headswitching noise varies a lot between VCR brands and I doubt if you could come up with one algorithm for all brands.
The result is far from perfect though but I do consider it an improvement.

The JVC 9600 PAL S-video signal has 2 bad lines at the top and 13 at the bottom. I suppose the whole image was digitized one line too low, omitting the 1/3 line at the top and adding a black line at the bottom.
My routine copies color of line 3 to line 1 and color of line 4 to line 2 (1-based numbering). Also 50% of brightness of line 1 is taken from line 3. No time base problems here.
At the bottom the right parts of bottom line 564 and 565 have shifted to the right, lines 566-574 have entirely shifted a few pixels, line 575 has the last 1/3 line of video and line 576 is black.
So my algorithm tries to find the best break points in lines 564 and 565 and best amount to slide the right parts to match with the line above them. For lines 566-575 I just look for the optimal shift.
Finally I paint over a few spurious blue spots at the bottom.

Of course this approach may result in diagonal objects ending in a vertical orientation at the bottom of the image. And sometimes a set of bottom lines get shifted by the wrong amount resulting in a more visible artifact.

Jan E.
jeschot
Site Admin
 
Posts: 694
Joined: 20 Feb 2008, 22:49


Return to JES Video Cleaner

Who is online

Users browsing this forum: No registered users and 1 guest

cron