Nonlocal Diffusion Tensor for Visual Saliency Detection

In this paper, visual attention transfer is formulated as a nonlocal diffusion equation. Different from the other diffusion based method, a nonlocal diffusion tensor is introduced to consider both the diffusion strength and direction. Along with the principle direction, the diffusion should be suppressed to preserve the dissimilarity between the foreground and background, and in other directions, the diffusion should be boosted to combine the similar regions and highlight the saliency object as a whole. Through a two-stages diffusion, the final saliency map is obtained and quantitative and visual comparisons are executed on two large benchmark databases. Experimental results demonstrate the superior performance of our method.