ipandzic++at++cui.unige.ch
Wed, 20 Sep 1995 07:05:36 +0200
> We are doing an experiment with placing video images of people in VRs.
> We are collecting the live image with a Indy (by the way, the Indy
> camera is just about junk, sorry SGI folks).
> What we would like to do is separate out the "talking head" from the
> background without using something like chroma-key, since we hope that
> people can use the technique in their offices (so we want to advoid
> placing a blue curtain behind people).
> We are having problems with our simple algorithm of sampling the
> background and then doing a delta when the "talking head" starts to
> get the pixels that are different.
> Just wondering if any of you have already tried this and have some
> suggestions on the best way to cut out the person from the background.
> thanks, Kim.
Hi,
We are about the same thing in a quite simple way - just take an image
of the background WITHOUT the talking head and store it in memory. Then
in each frame we compare the backgroung with the current frame and the
head is where we find a significant difference. This is simple to implement
and works not-too-badly. It requires a static backgroung, i.e. if a
colegue is working behind you in the camera field of view it won't work.
Hope this helps,
Igor
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Igor Pandzic ipandzic++at++cui.unige.ch
MIRALab
University of Geneva tel. +41/22/705 76 56
24, rue General-Dufour 705 77 63
CH-1211 GENEVE 4
SWITZERLAND fax +41/22/705 77 80
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:51:53 PDT