摘要
In this paper, we propose a new computational model of visual attention based on the relevant characteristics of the Human Visual System (HVS) and sparse features. The input image is first divided into small image patches. Then the sparse features of each patch are extracted based on the learned independent components. The human visual acuity is adopted in calculation of the center-surround differences between image patches for saliency extraction. We choose the neighboring patches for center-surround difference calculation based on the relevant characteristics of the HVS. Furthermore, the center-bias factor is adopted to enhance the saliency map. Experimental results show that the proposed saliency detection model achieves better performance than the relevant existing ones on a large public image database with ground truth.