A new algorithm combined fractal signiture and gray  correlation for exemplar matching

The exemplar matching makes great sense in the analysis of image, acquisition of the information and target recognition etc.. For does not need to extract features firstly, algorithm based on region and texture becomes a hot issue. But the computational complexity of this kind algorithm is high. To solve the problem, a algorithm combined fractal signiture and gray correlation for exemplar registration is proposed in this paper. It is divided into three steps: Firstly, compute the fractal signatures of the pixels of the image, and gain the fractal signature of the exemplar according to its position in the image. Then, find the regions that maybe the same or similary to the exemplar image using the fractal signature. Finally, eliminate the false matching regions by use of gray correlation, and gain the final matching result. In addition, a fast method for computing fractal signature is also propose in the paper to improve the efficiency. Experiments show that the propose algorithm is effective and potent