A Robust Object Tracking Method for Noisy Video using Rough Entropy in Wavelet Domain

In this paper we have proposed a robust object tracking method using rough entropy and flux in wavelet domain. The tracking framework necessitates robust and efficient but accurate methods for segmentation and matching. The object is represented in wavelet domain features to minimize the effect of frame to frame variations and noise. The concept of maximizing rough entropy in wavelet domain helps in finding out the threshold value to make a distinction between the object and the background pixels in a vague situation. The search for the candidate subframe is made fast by using the motion prediction algorithm. A measure based on flux in wavelet domain combined with the number of pixels in the object has been developed. The proposed tracking algorithm yields better results even in noisy video as shown in the experiments. The results show that the wavelet domain segmentation and tracking improves the localization error approximately by 5-7%.