Development a new technique of ROI and feature extraction method for palmprint image based biometric

Human recognition and verification based on palmprint image is an important technology and it has attracted much attention. The palmprint image consists of principle curves and wrinkles and also important texture such as miniscule points, so the system can achieve a high accuracy because of available rich information in palmprint. Palmprint is the impression made by the inner surface of the hand below the fingers to the wrist. It is one of the human physiological characteristics and it is universal, unique, permanent, collectible, consistent, comparable, inimitable, and tamper-resistant. It universal because every person has a palmprint and it is unique because every palmprint is different from other person's palmprint; even identical twins have different palmprint features. Palmprint is easy to collect and consistent because it does not change much with time and it's hard to imitate because of its size. It’s tamper-resistant because it cannot be changed and hiding it is difficult. This work aims to set an objective for the pre-processing step and feature extraction for palmprint biometric. A new ROI cropping technique is proposed in this work to overcome the rotation and translation issues in a pre-processing stage, firstly used Enhanced Correlation Coefficient (ECC) which is recommended to assess the parameters of the motion model. This approach has two advantages, firstly In contrast to the conventional similarity measure of differences in pixel intensities, ECC remains unaffected or invariant to photometric distortions about brightness and contrast, secondly, Moore’s neighbour boundary tracing algorithm is used to extract the boundary of the palm contour. Following this, the distance between the boundary points on the fingers side and the center of the image is calculated, based on the x-axis coordinate. Finally, local minima are extracted from the distance values to construct a cropping coordinate system. In the feature extraction stage since local binary patterns (LBP) is texture based descriptor but the size of the features increases exponentially with the increase in the number of neighbour which leads to an increase of computational complexity and gave poor accuracy .In this proposed work vertical local binary patterns (VR- LBP) is used, nevertheless, KAZE feature detector has been used to detect the most strong point in the palm image after using VR-LBP to ensure the accuracy of the palmprint biometric system. Last and not least, Principal Component Analysis (PCA) is used for dimensionality reduction which is way to reduce the complexity of a system and avoid overfitting .These methods applied on PolyU database and the IIT Delhi touchless palmprint database and for matching stage the Euclidean distance classifier is used and the system achieves a 99.7%, for the verification result the system has been achieved 98.78 %.