Fast object tracking with long-term occlusions handling in dynamic scenes
Mohammad Amin Bagherzadeh, Mehran Yazdi · 2014
In this paper, we present a simple yet fast and robust long-term tracking algorithm of arbitrary objects, where the object may become occluded or leave-the-view in a video stream, which exploits the Mean-Shift (MS), appearance model and saliency map for visual tracking. The Fast Fourier Transform is adopted for saliency detection in this work. The proposed Mean-Shift and Saliency Detection Tracker (MSDT) algorithm runs in real-time and numerous experimental results on several challenging image sequences demonstrate that the proposed tracking framework more favorable performance than the state-of-the-art methods in terms of accuracy, efficiency and robustness.