Saliency detection via statistical nonredundancy

A novel algorithm based on statistical non-redundancy is pro-posed for saliency detection in natural images. By modeling site neighbourhoods as realizations of other site neighbour-hoods under a Gaussian process, the saliency of any arbitrary site can be characterized by the statistical non-redundancy of its site neighbourhood with respect to other site neighbour-hoods in a given image. Preliminary results using natural im-ages show that the proposed method provides improved pre-cision vs. recall characteristics over previous methods such as spectral residuals and spectral whitening. 1