Integrating Perceptual Properties of the HVS into the Computational Model of Visual Attention

The integration of bottom-up, image-driven cues and perceptual properties of the human visual system (HVS) is crucial for the selection of salient locations. The previous studies have mainly focused on the bottom-up cues. In this paper, we propose a new model that takes both two factors into consideration. More specifically, two bottom-up conspicuity maps for early-visual features (i.e., intensity and orientation) are modulated by the value of visual contrast sensitivity of every scene location and are then combined with the color conspicuity map to create an overall perceived saliency map such that this model's predictions agree well with outputs of our visual perception.