A New Fast Matching Algorithm by Trans-Scale Search for Remote Sensing Image

A new fast matching algorithm for remote sensing images is proposed. The algorithm adopts a coarse to fine matching process. A remote sensing image is decomposed into multi-scale images which form a wavelet image pyramid by the way of DWT (Discrete wavelet transform). Extract a low frequency sub-image from the wavelet image pyramid to carry out a rough matching operation between the sub-image and target image, and then get a rough position after cluster analysis. Get a suitable position with center at the rough position from remote sensing image to accomplish fine matching operation. The algorithm has been analyzed theoretically from the signal processing point of view, and a sufficient condition is given to select wavelet filter. Simulation results testify that proposed algorithm not only distinguishes remote sensing images precisely, but also cuts down matching time greatly. It is only 25.04% of SIFT algorithm, and 35.36% of SURF algorithm for matching time.