Modeling the masking effect of the human visual system with visual attention model

It is well known that the human visual system (HVS) cannot sense all changes in an image/video due to its underlying physiological and psychological mechanisms. We propose a complete masking estimation model for image/video in this paper. In our model, a very important mechanism of the HVS, visual attention, is incorporated to the existing just noticeable difference (JND) estimation models. A formula for evaluating the weighting map is adopted to deal with the effect of visual attention using foveation technique. Experimental results with noise shaping and subjective viewing confirm the improved accuracy of the proposed model over the existing relevant models.