A visual saliency detection algorithm based on the image anisotropic in the frequency domain

A frequency domain algorithm using the image anisotropic is proposed for visual saliency in this paper. Based on the SSS algorithm, we generate 8 saliency maps corresponding 8 Gaussian kernels. Then, use the Renyi entropy to divide these saliency maps into two classes and select one optimal map for each class. Finally, combine these two optimal maps for the finally saliency map. The experimental comparison between the proposed and original approach demonstrate the effectiveness and accuracy of our method.