Efficient object tracking in video sequences by means of LS-N-IPS

A recently introduced particle filtering method, called LS-N-IPS (local search N-interacting particle system), is considered for tracking objects on video sequences. LS-N-IPS is a computationally efficient particle filter that performs better than the standard N-IPS particle filter, when observations are highly peaky, as is the case of visual object tracking problems with good observational models. An implementation of LS-N-IPS that uses B-spline based contour models is proposed and is shown to perform very well as compared to similar state-of-the art tracking algorithms.