Subject: Re: [MPlayer-users] show interlaced/progressive info..

[rbtq7ol02 at sneakemail.com]

> On Fri, 14 Nov 2003, Tuukka Toivonen wrote:
>
> One idea I wonder if you have tried it: compute correlation first
> between vertically adjacent scanlines, the compute correlation between
> every other vertically adjacent scanline (leaving one scanline between)
> and the compare the average correlation. If it is larger for scanlines
> which have one scanline between them, then the video is likely to be
> interlaced.

Erm...  this sounds very good.  Also potentially very fast, in MMX (or
even better in SSE, with psadb).  My math says that keeping an
accumulation of these differences over the whole image, even for insanely
large images, is plausible, too (accumulated absolute difference should
be well under 2^32, even for insane cases like 1920x1080 in RGB).  Mind
if I try out this test for something else I'm working on for MythTV?
-- 
  Andrew Mahone
  andrewmahone AT eml DOT cc

-- 
http://www.fastmail.fm - The professional email service