Research On Second - Order Differential Feature Extraction And Face Recognition Of Image

Face recognition is an important technology for identity authentication. Senior emerging technology is developing rapidly in the information age. The traditional authentication methods could not meet the needs of the society, Effective biometric authentication technology has become the mainstream of the market. Face recognition is one of biometric authentication technology, It has the advantages of easy acquisition and non-contact, and can be accepted by the public easily. How to improve the rate of facial image recognition is the key of this technology.This thesis presents a new algorithm for texture feature extract that based on second-differential, apply it to the face image recognition. Firstly, the thesis introduces the principle of traditional face recognition algorithm.Analysis of shortcomings of basic local binary pattern. Then,propose some new algorithms and improved algorithms through combining with the second-differential and center-symmetric theory.Lastly,simulate a variety of identification algorithms by using standard face database on the platform of MATLAB, and analysis the advantages and disadvantages of the algorithms.The main innovation of this paper:1.The CS-LDP algorithm based on multichannel Gabor filter is proposed.Gabor has the characteristics of multi-scale and multi direction.These properties can be used to extract CS-LDP feature of images in several directions;2.The LCCP algorithm is proposed. This algorithm uses second-differential principle to extract the texture feature of images fully, broke through the limitations of the first order differential extraction of texture information.And an improved algorithm based on LCCP is proposed.It expands the neighborhood pixels of LCCP algorithm, and increases direction of feature extraction;3.Fusion of LCCP and CS-LDP algorithm.Fusion of convex and concave feature and gradient feature, then it is represented by a feature vector.The experimental results show that the new algorithm can extract more feature information, enhance the robustness of face recognition.The traditional face recognition pattern is built based on the first order differential algorithm, information co...The second-differential algorithm can overcome the change of illumination, posture, facial expression and other factors that influence on recognition rate more efficiently.Because the main organs that face contains,such as eyes, mouth, nose and so on have the concave convex characteristic.The second-differential algorithms can extract the convex and concave feature information of images.At the same time, this thesis briefly introduces the central symmetry theory and interpolation principle, not only insert new pixels but also link field pixels. Improves the recognition rate of face images.