Based On A Non-binding Research Of Face Recognition Method

As one of the most important biometric techniques, face recognition implements identity authentication using the features extracted from human face images, which possesses advantages of being natural, simple, intuitive and passive over other biometric techniques such as fingerprint recognition and iris recognition. Therefore, face recognition has a broad application prospect in information security, criminal detection, access door control etc.Although face recognition has been actively studied over the past decade, the state-of-the-art recognition systems yield a satisfactory performance only under controlled scenarios, but their performances degraded significantly under unconstrained situations due to the complexity of the human face and the environment, such as changes in facial expressions, posture, the image environment light intensity conditions. In addition, face obstacles (glasses, beard) will greatly affect the performance of these face recognition methods. Therefore, the face recognition under unconstrained condi-tions remains a challenging proplems. This thesis addresses this issue and proposes several unconstrained face recognition approaches. The main contributions of this thesis are as follows:1. A novel method for face recognition based on modified pulse-coupled neural network is proposed. In this work, facial images are decomposed into a sequence of binary images using a modified pulse-coupled neural network (M-PCNN), and then the information entropies of each binary image are calculated and regarded as features. A support vector machine-based classifier is employed to implement recognition and classification. It overcomes the disadvantage of standard PCNN model with high number of parameters. Theoretical and experimental results show that the proposed approach is robust to the variations of orientation, position and illumination conditions in comparison with the recent methods.2. Face recognition based on2D-Gabor faces and PCNN.2D-Gabor wavelet transform decompose the image in different directions and different scales which can capture image spatial frequency, spatial location and direction information. These features is not sensitive to light conditions and face gestures. This paper describes a novel method for face recognition combining Gabor wavelet transform and Pulse Coupled Neural Network (PCNN). In the proposed algorithm, facial images are decom-posed by2D Gabor filters. Each Gabor subband is decomposed into a sequence of binary images using a pulse-coupled neural network, and then the information entropy of each binary image is calculated and regarded as a feature. A support vector machine-based classifier is employed to implement recognition and classifcation. Theoretical and experimental results show that the proposed approach is effectiveness in comparison with the recent methods due to the fact that2D-Gabor wavelet not sensitive to the change of light conditions..3. Face recognition based on NSCT and PCNN. A facial feature extraction algorithm based on Nonsubsampled Contourtlet Transform (NSCT) and Modified Pulse-Coupled Neural Network (M-PCNN) is proposed. By using NSCT, the input images are decomposed into a number of sub-images with various scales and directional features. Then, the different subbands are decomposed into a sequence of binary images using M-PCNN, and then the information entropy of each binary image is calculated and regarded as a facial feature. A support vector machine classifier is employed to implement recognition and classification. Theoretical and experimental results show that the proposed approach can adapt to the variations of illumination and pose in comparison with the recent methods, and yields a better performance in terms of accuracy of face recognition under the unconstrained conditions.