Palmprint Image Feature Extraction Methods Research

With the development of technology, the conventional personal authentication approaches expose their drawbacks. The password based methods are forgettable and easy to be cracked. While the ID cards based methods are easy to be counterfeited, lost and shared. The aforementioned problems make the personal authentication unsafe to some extent. Meanwhile, the personal information or some specific authorities may encounter with the risk of be used. Compared with the passwords or ID cards based solutions, biometrics authentication is much more preferable and reliable, and plays an important role in the personal authentication.Biometrics is composed of face, ear, fingerprint, palmprint and other inherent characteristics of human. The face and fingerprint have extensive applications due to their high acceptance rate. Palmprint is a relatively new but a later member of biometric characteristics. It contains a rich amount of stable texture features, which lead to very high recognition accuracy. Meanwhile, the merits such as low-cost, user friendliness and high matching speed make it practicable to use in a large scale. Many state-of-the-art palmprint texture and orientation based feature extraction methods such as Palm Code, Fusion Code, Competitive Code, RLOC have been proposed.The main work of this paper is as follows:1. Reviewing the researching status of the biometrics and palmprint based verification, studying some state-of-the-art palmprint feature extraction and matching approaches and coding them as well.2. Introducing the preprocessing stages of palmprint image in detail, and highlighting the importance of smooth denoising and morphological operations on palmprint image preprocessing.3. Proposing an extension of Orthogonal Code and Fusion Code method, namely Orthogonal Fusion Code (OFC) method. The experimental results show the superiority of the proposed method to the original two methods.4. Performing Linear Discrimination Analysis (LDA) on the data to achieve feature dimension reduction in advance. Then the Sparse Preserving Projection (SPP) method first proposed by Qiao et al. and used in face feature extraction is applied to the preprocessing data to further extract features. The experimental results demonstrate that the above scheme outperform the proposed LDA,LPP methods in some cases. Coding the SMCC method and comparing it with the Competitive Code. The experimental results show that the sparse representation has a wide room to study.