Visual saliency is a type of visual feature simulating visual attention selection mechanism in biological system

and has better robustness and invariance. A rapid scene registration method based on visual saliency is proposed in this paper. Firstly

the method adopts amplitude modulation fourier transform to construct saliency map; then rough location，pre-selection and reliability order of salient points are achieved by analyzing the extremal properties and information richness of saliency feature; morphological operation is used for salient scene region growing and merging on the image. On this basis

SSIFT algorithm is proposed

which can largely reduce the computational cost of scene registration. University of Southern California scene dataset is used to test the presented method

the experimental results indicate that the method has good invariance under image scalings

rotations

translations and lighting variations. Compared with classical SIFT algorithm

the method works at higher speed and has higher recognition rate.