Foreground/background difference saliency via information maximization for moving target detection

This paper proposes a novel method for moving target detection using foreground/background difference saliency via information maximization. Information maximization saliency map of the current frame is first generated to highlight the moving targets. To reduce the negative effects from the clutter scene, saliency map recording the interference factors is also constructed for a static background, and the moving target is detected based on the difference saliency, rather than the traditional background subtraction. In order to calculate better basis functions for saliency, we update constantly training samples for basis functions using background salient regions. The experimental results of our method are very encouraging, especially in the complex scene with many camera jitters and background disturbances etc.