Fingerprint image is converted into the spatial frequency domain using Fast Fourier Transform. The logarithmic operations are applied on the resulting image. Then the image is normalized. The polar coordinates are converted into Cartesian coordinates. The inner energy, the outer energy and the overall energy from the spectral image are calculated and considered as feature vector.paragraph{Statistics}The gray level value of fingerprint pixels can be analyzed to detect liveness property of the fingerprint.
The gray level distributions is modelled as first order statistics. The first order statistics are energy,entropy, kurtosis, variance, kurtosis, skewness, and coefficient of variationA new method by combining ridge signal and valley noise analysis is proposed for anti-spoofing in fingerprint sensorscite{c38}.This method quantifies perspiration patterns along ridges in live subjects and noise patterns along the valleys in spoofs. The signals representing gray level patterns along ridges and valleys are explored in spatial, frequency and wavelet domains. Based on these features, liveness detection is performed using standard pattern classification tools such as neural networks and classification trees. A new liveness detection method based on noise analysis along the valleys in the ridge-valley structure of fingerprint images is given.
Unlike live fingers which hold a clear ridge-valley structure, artificial fingers have a distinct noise distribution due to the material properties when placed on a fingerprint scanner. Statistical features are extracted in multi resolution scales using wavelet decomposition technique. Based on these features, liveness separation is performed using classification trees and neural networks. paragraph{Skin elasticity}Distortions due to the rotation and pressure of the finger on a sensor result in different elastic characteristics of the materials. Liveness can be detected by comparing these distortions through static features. The elastic deformation due to the contact of the fingertip with a plane surface was analyzed in cite{c26} since a fake fingerprint presents different deformations than a hot one. The elastic behavior of a live and a fake finger was analyzed by using a mathematical model relying on the extraction of a specific and ordered set of minutiae points.
Fake fingerprint detection using skin distortion is discussed. The user is required to move the finger while pressing it against the scanner surface, to produce skin distortion. When a real finger moves on a scanner surface, it produces a significant amount of distortion, which can be observed to be quite different from that produced by fake fingers.
Usually, fake fingers are more rigid than the skin and the distortion is definitely lower.paragraph{Thin-plate Spline:}Thin-plate spline model is used to analyze the finger distortion. For the fingerprint image, the bending energy is computed.
The similarity of the bending energy is calculated.paragraph{Optical Flow:}The user moves the finger on the fingerprint scanner surface either in clockwise or counter-clockwise direction which produces the skin distortion. A sequence of frames is captured during the movement and specific features related to skin distortion are extracted. Then optical flow between consecutive frames, distortion map and distortion code are evaluated.