I use Kalman filter for background/foreground separation as discussed in "Adaptive Background Estimation and Foreground Detection using Kalman-Filtering ". Basically is implementing the following equations.
Where S-hat is estimated pixel background intensity. I have already implemented the above equations. I see that S-hat is following/close to the input image intensity value at s(x,y,ti), that means I assume, the implementation is correct. In that case, how to extract the foreground mask? Thanks
This question is answered perfectly well in the paper.
You find the map by checking if s
and s-hat
are close to each other. The below part of the paper has all the equations and a bit of commentary on how to choose the threshold.