Scale Invariant Kernelized Correlation Filter Based On Gaussian Output

Kernelized Correlation Filter (KCF) is one of state-of-the-art trackers. However, KCF suffers from the drifting problem due to inaccurate localization caused by the scale variation and wrong candidate selection. In this paper, we propose a new method, named Scale Invariant KCF (SIKCF), which estimates an accurate scale and models the distribution of correlation response to address the template drifting problem. The features of SIKCF consist in: (1) A scale estimation method is used to find an accurate candidate. (2) The correlation response of the target image is reasonably considered to follow a Gaussian distribution, which is used to select the better candidate in tracking procedure. Extensive experiments on the commonly used tracking benchmark show that the proposed method significantly improves the performance of KCF, and achieves a better performance than state-of-the-art trackers.