Improved kernelized correlation filter tracking by using spatial regularization

The correlation filter based trackers have drawn much attention due to their encouraging performance on precision, robustness and speed. In this paper, we introduce the spatial regularization component into the ridge regression model used by classical kernelized correlation filter (KCF) to improve its performance. It overcomes the fact that the traditional KCF does not consider the prior spatial constraint of the feature distribution of the target. We found that, after adding the spatial regularized function, we can solve the ridge regression formula efficiently with the property of circulant matrices. In this way, we can simultaneously keep the realtime and improve the tracking performance. Finally, we evaluate the proposed SRKCF tracker on the OTB-2013 and OTB-2015 comparing with 36 trackers and our tracker achieves state-of-art. Comparing with the SRDCF which applies the spatial regularized function, our algorithm achieves comparable performance with the obvious advantages in speed.