Background Subtraction for short videos

Question

I was going through this paper: Adaptive Background Mixture Models for Real-Time Tracking(1999) by Chris Stauffer, W. Eric L. Grimson

Their method is ok for a long video where one can start making blobs after some time when the gaussian mixture models gets stabilized.

But this won't apply for short videos e.g. http://cvrc.ece.utexas.edu/SDHA2010/videos/human_interaction.zip Here from the first frame itself, some portion is actually part of foreground.

Can anyone point out me some paper(s) where this problem has been handled?

Thanks, Kaushik

I had the same problem. What I did was run the video through the algorithm once to initialize the background model, then a second time (retaining the initialized background model) to do the segmentation. For off-line segmentation this is fine. — jilles de wit, May 02 '11 at 09:15

score 0 · Answer 1 · answered May 03 '11 at 11:50

0

Monocular Video Foreground/Background Segmentation by Tracking Spatial-Color Gaussian Mixture Models looks like a really fast-to-initialize method. I'm not sure it can achieve segmentation on the first frame though.

answered May 03 '11 at 11:50

jilles de wit

score 0 · Answer 2 · edited May 23 '17 at 10:34

0

Statistical Background Subtraction for a Mobile Observer (Section 4) also handles this issue of background not yet confirmed. Facing some issues in understanding.

edited May 23 '17 at 10:34

Community

answered May 07 '11 at 18:01

Kaushik Acharya

2 Answers2