An Application of an EM Algorithm for Skew Detection of Signatures in Text Images

For security purposes of important documents and transactions in real world applications, we generally use biometric techniques for the authentication and validation of a person. If one has to achieve accurate results in the identification and verification process using a signature in text images as a biometric trait, we need to remove the skew of the signature in text images. In the preprocessing stage many phases are being carried out, among these phases, the signature in the text image, skew detection is the most significant phase, because these deskewed results will be used as one of the features in the feature extraction phase to identify and verify the signature. In this article we are proposing a novel method for skew detection of the signatures in text images using an estimation and maximization (EM) algorithm which is efficient and fast. The EM algorithm sequentially works in two stages, the combination of estimation (E-step) and the maximization (M-step) which helps in detection of the skew in skewed signatures in text image accurately.