Visual tracking using multi-channel correlation filters

Tracking-by-detection methods are widely used in video based object tracking. The correlation filters, which use Gaussian function as output response and train the filters in Fourier domain, provide excellent tracking performance and high possessing speed. However, the classical correlation filter is not so robust in practice as it uses linear classifier and processes raw image pixels. In this paper, we extend the linear correlation filter to multi-channel case, which can incorporate multiple feature channel descriptors into the processing so that the robustness of filter could significantly be improved. In our demonstrating system, the multi-channel HOG descriptors are utilized to represent the image patch. Experimental results show that the proposed method outperforms state of the art trackers like MOSSE and CSK.