Online robust object tracking via a sample-based dynamic dictionary
Yang Liu, Yibo Li · 2013
We propose an online robust object tracking algorithm based on a sample-based dictionary. The sample-based dictionary in our method means that the over-completely dictionary of sparse coding algorithm is formed by using the sample basis extracted from video images. Different from the other tracking methods that use the object features and a set of boosted classifiers, the proposed algorithm considers the raw image patches around the object as basis vectors and the maximum a posteriori is used to decide the object position in the next frame. Our method is simple for the dictionary that is updated automatically and the object is updated to alleviate the visual drift problem in every frame without learning process, which needs much time consuming. Experiments are conducted on the challenging sequences to demonstrate that the proposed method is fast and effective. The results show that the proposed method outperforms the current state-of-the-art methods.