Signature Identification Based on Pixel Distribution Probability and Mean Similarity Measure with Concentric Circle Segmentation

This paper studied a new method of signature identification. First, the center of image of signature has to be found. This center is regarded as the center of a circle. The radius of circle is the greatest length from the center to the edge of signature. The radius is divided into 10 equal-length parts. According to 10 different radiuses, the signature image is segmented into 10 sections by 10 equal-interval concentric circles. 4 segmentation modes can be made with the combination of 10 different-radius circles. Then, the pixel distribution probability of signature in every section is calculated. It is the ratio of the pixels of signature of every section to the pixels of signature. It represents the structural and statistic feature of signature. Finally, the mean similarity measure based on the distance and the correlation coefficient of pixel distribution probability between identified signature and standard signature can be calculated. According to the mean similarity measure and the statistical principle, the result of signature identification can be got. By experiment of 360 signatures identification, the accurate rate based on above method is more than 98%.