Biometric Image Processing

An overview of biometric systems has been presented. For decades, many government and public establishments have used biometric authentication for access control. Today, the primary application of biometrics is shifting from the physical security, where the access to the specific locations is usually monitored using standard security identification mechanisms such as ID or token-based mechanisms in combination with fingerprint biometric, to remote security where the methods of crowd monitoring using video surveillance take advantage of gait biometric, for example. The popularity of such approaches has increased dramatically as the new technological devices are coming on the market every week, capability to process massive amount of data is doubling every few months, and   algorithm development by the leading IT companies and the universities research centers is tripled in the last years.

          From the gamut of research on biometric authentication, we observe that the overwhelming part of biometric data processing is realized by using image processing and pattern recognition methods and algorithms. As the mainstream direction of biometric image processing, appearance-based methods extract biometric features from the row image by analyzing appearance of the whole image as an entity or a vector in a high-dimensional image space. Such factors as color scheme, orientation, background, luminance, saturation are being analyzed and processed either pixel by pixel or through projection on sub-spaces  such as in Principal Component Analysis (PCA) methods. As the most evident examples, we consider face, iris and ear biometrics in this context.

The goal of any intelligent processing is to minimize overhead associated with performing computations while at the same time to maximize an output. The same principle governs behavior of most public and commercial organizations—to achieve high production by resource and processes optimization. While appearance-based methods excel in capturing even subtle features in the multitude of high-dimensional data, sometimes generalizing the results and noting common patterns leads to process optimization without sacrificing the security system performance. This section presents topology-based methods, which work best with biometric data that has prominent geometric features, such as fingerprint or hand/palm biometrics. We start by outlining the topology-based methodology with the roots in computational geometry.