METHOD FOR MEASURING SIMILARITY USING FRENQUENCY RANGE

摘要

A method for measuring similarity of paper money is provided to prevent detection error of a concavo-convex part damaged by an edge search. An edge element about the binary paper money image is extracted. The rotation angle of the paper money image is obtained for correcting the rotated paper money. The original image is pivotally transformed by using the calculated rotation angle. The binarization is performed through the dynamic standard value of the luminance of the paper money part In order to extract the paper money part to be processed. In the binarized X,Y project of the image, the paper money about the size of the width and the length are obtained and the center of the square zone is obtained. The frequency information of compressed image is extracted by using the fast Fourier transform algorithm. The similarity computation is performed by comparing the spatial frequency domain. The paper money is classified by the reciprocity similarity calculation of the spatial frequency feature based on the spatial frequency feature data base which is prepared from the extracted spatial frequency.