Article Preview
Top1. Introduction
Biometric means measuring the uniqueness of a person and it is the measurement of one’s life. In biometrics physiological characteristics are being used to identify an individual. Person’s authentication is done in various real world applications like- voter ID, bank cheque books, passport, adhar card, pan card, credit and debit card transactions and document authentication etc, for these applications we generally use various biometrics techniques such as recognizing the face, fingerprint verification, voice recognition and verification, handwriting classification, iris recognition and many more are being used for the identification of a particular person in real world applications (Sabourin et al., 2012). A well-known image analysis product is the Optical Character Recognition (OCR) software that recognizes character in a scanner document. Signature in text images identification and verification is one such example of behavioral biometric application (Kasturi et al., 2002). Signature in text images identification and verification (SIV) system is one such example of behavioral biometric application. Every signature in text images signed by individual will be unique and cannot be easily copied. In order to authenticate a person, classification is the most important step to be done in this system very accurately and effectively. Two main steps we can find in classification i.e., identification of a signature in text images and verification of a signature in text images. First is to identify a person and second is to verify a person. It is essential to have a computerized SIV system to find the fraudulence done in signature in text images forgery domain. In most of the offline SIV systems dynamic features of the signature in text images are not available, but we can save the image of the signature in text images in digital format and can apply the digital image processing techniques using various algorithms to make the SIV system effective and accurate (Rasheed, n.d.).
In order to make this SIV system more effective, first we should do is to preprocess the signature in text images. Preprocessing of a signature in text images is a must and necessary step to enhance the selected features like skew, size and to remove the unwanted information to improve the quality of the image. Based on the accurate results of preprocessing stage we extract the features of the signature in text images for SIV system (Kanawade & Katariya, 2013). The preprocessing step is functional in both training and testing fields. Most of the document image preprocessing systems will face problems like unwanted data (noise), which makes the signature in text images lines and curves to touch in between them, pixels that are so out-of-the-way and the tarnished images. Skewed images give less information while extracting the information of the signature in text images. Images will be in different sizes based on the signer this will result in getting less accuracy. These tribulations may cause high level distortions in the digital signature in text images image and hence effect in indefinite features and on the contrary gives poor recognition and verification rate. Skew detection phase is the main phase in the preprocessing stage and which plays a vital role in giving the exact results to extract the features in feature extraction stage and to identify and verify the signature in text images in SIV system. Skew in images occurs when the image is tilt or when the signer has signed the signature crossly. These things effects on prediction of the signature in SIV sytem. There have been significant of work is done in this area; still we are unable to get accurate results which consequences in getting the false results. To trounce from these tribulations, we came up with a new method called skew detection using EM algorithm. Authors in their paper (Gaurav et al., 2018; Jena et al., 2019; Ramappa et al., 2012) have explained about EM algorithm briefly. Estimation step is used to predict and estimate the amount of skew present in the signature. Maximization step helps in correcting or deskew the skew present in the signature image.