An Improved Mean - Shift Algorithm with Self-Scaling Tracking Window
Priya K. Mahajan, Mugdha M. Dewasthale · 2017
In standard mean shift based object tracking algorithm, kernel bandwidth (BW) is fixed due to which object tracking fails if object is changed in the size. To overcome this problem, this paper proposes the method to increase tracking potential of the standard Mean-Shift algorithm (MSA). Firstly, it calculates the size of the object, if it changes then finds the size of kernel-BW. Accordingly, it adjusts the kernel-BW of the MSA. Finally, using improved MSA, the tracking object has been accurately located with proper kernel-BW. The experimental result of the proposed algorithms has been proved that it is improved than the existing standard mean shift algorithm.