Source Printer Authentication for Printed Documents Based on Factor Analysis

Source printer authentication for printed documents is very hard in printed documents identification. When the characters in samples and test material are completely different, the problem becomes extremely challenging. In this paper, a new method primarily based on factor analysis has been proposed to solve the problem. Concretely, we utilize bilinear model to analysis feature matrix to separate the printer factor and text factor entirely. Then, we extract text-independent printer features of the samples approximately. Finally, expectation maximization algorithms are utilized to obtain maximum after posterior probability to classify. Experimental results show that our proposed method was effective to improve the recognition rate.