Moving Object Detection Using Background Subtraction in Wavelet Domain

Detecting and tracking of any suspicious moving objects is of the prior importance in video surveillance. In this paper, the background subtraction method has been employed in the wavelet transform domain. By masking the approximate wavelet coefficient to zero, the detailed coefficients provide the necessary information about the moving objects only, thereby reduces the probability of misdetection and overcome the illumination variations. The performance of transform domain method has been compared with popular three-frame differencing method, both employing background subtraction, global thresholding and morphological dilation. The proposed method is relatively superior to that of the later case on visual inspection.