Abstract: In order to determine the detection threshold in moving objects extraction, an adaptive detection threshold determination method based on the two-dimensional spatio-temporal entropic principle is proposed. The motion detection is reduced to two binary partition problems, and two-dimensional entropic principle is used to determine the threshold vector(S,T), where the explicit function form and parameters of background distribution are not needed to be known. A fast algorithm for calculating the two-dimensional entropic threshold is put forward, in which repeated calculation is avoided by using the integral and iterative operation, thus reducing the computational complexity from O(N4) to O(N2) for an N gray levels image. The experimental results show that the information of moving objects which have a low contrast to background can also be extracted completely by this method in real time.

Key words: two-dimensional entropic principle, motion detection threshold, moving object extraction, integral operation