Study On Algorithms Of Face Recognition

Face Recognition, which has been studied for more than 30 years, is one of the challenging researches in pattern recognition and machine vision, it can be widely applied in the public security, information security and human-computer interaction. By now, face recognition technology under well-controlled imaging condition is practically usable, while the performance of algorithm dramatically decreases under uncontrolled environment, which involves the variation of head poses and illumination et al. Face recognition with pose variation are studied in this thesis, and in these theories the most direct algorithm is generating frontal view face image using linear object-class theory and pose transform matrix which not only improves the face recognition rate but also avoids the complicated 3D model.The thesis studies the correlative issues based on the former works. The main points are as follows:(1) An improved composing frontal face image method is proposed using independent component analysis (ICA) algorithm. Principal component analysis (PCA) just uses the data's second-order statistical information, while ICA uses the high-order statistics besides the second-order statistics, so the composed quality and face recognition rate are improved.(2) Because the method could compare and classify directly using the composing frontal faces'eigen coefficient, the combination algorithm of wavelet transform and ICA and the combination algorithm of genetic algorithm (GA) and ICA are proposed to study the multi-pose face recognition. The wavelet transform wipes off the redundancies and reduces the operation complication without influencing the recognition rate. The second new method could get the optimization cut subspace combination using the GA to optimize the different pose ICA eigen subspace, so the method could get the better face recognition rate.(3) When the single image is magnified using the traditional interpolation algorithm, the image's high frequency details are dropped because of being restricted by the information and the interpolation effect isn't obvious. The paper applies the frontal face composing algorithm to reconstruction of super-resolution face images and receives the famous result.