A salient object detection framework beyond top-down and bottom-up mechanism

Traditional saliency-based attention theory supposed that bottom-up and top-down factors combine to direct attentional behavior. This dichotomy fails to explain a growing number of cases in which neither bottom-up nor top-down can account for strong selection biases. Thus, the top-down versus bottom-up dichotomy is an inadequate taxonomy of attentional control. In this study, a general computational salient objects detection framework beyond top-down and bottom-up mechanism is presented. It possesses three parts: selection history, current goal and physical salience. Selection history is integrated with current goal and physical salience to compose an integrative framework. An image window saliency is defined as the objectness score of the window. Experimental results on challenging object detection datasets demonstrate that physical salience generates bottom-up saliency map for highlighting the salient regions of image, the main effect of selection history is to concentrate on salient objects, the current goal has strong effect to detect correct salient objects.