Kernel Subspace Integral Image Based Probabilistic Visual Object Tracking

This paper presents a novel object tracking algorithm. Object appearance and spatial information is learned from a single template using a non-linear subspace projection. A probabilistic search strategy, based on particle filter, is employed to find object region in each frame of the video sequence that best models the target object in the subspace representation. Particle filter estimates the posterior distribution using weighted samples. Increasing the number of samples increases the estimation accuracy at the cost of increased computations. We, therefore propose a novel kernel subspace integral image framework, which allows the tracker to densely sample the state space without loosing computational efficiency. The algorithm is tested on real world tracking examples to demonstrate the performance.