Facial Image Recognition Research On Illumination&Expression-Robust

Face recognition technology draws a widely public attention because of its non-contact, fast and easy-to-use features. And also a growing number of researchers engage in the work of research in the areas for its huge market in the security and other fields.This dissertation carries out related research focuses on Facial Image Recognition method on Illumination&Expression-Robust, the main research work is as follows:(1) Facial Image Recognition method on Illumination-RobustThe dissertation builds a method of facial Image Recognition on Illumination-Robust. The method first logarithmic enhances face images, then smoothes the image using the total variation model based on the gradient, and finally gets the available illumination normalized Logarithmic Total Variation Quotient Image.(2) Facial Image Recognition method on Expression-RobustThe research builds a method of facial Image Recognition on Express ion-Robust. The method creates a face geometric features model, and establishes weighted angle vector as several human face geometric eigenvectors. So you can set up the faces’ geometric feature space and textural feature space and complete the face identification work combines with the two eigenvectors. Using the method the recognition rate can reach94.6%in Yale databases. The experimental results show that this method can effectively eliminate the impact caused by expression in face recognition results.(3) Facial Image Recognition method on Illumination&Expression-RobustThe study presents a method of facial Image Recognition against Illumination&Expression-Robust. The method first enhances the contrast of the image using histogram equalization, and then sets up a local binary pattern map and regional gradient map and calculates matching cost between every two feature map by dense matching method, then finally completes face recognition work. This method can not only eliminate the effects of illumination and facial expression in face recognition results, but also can handle Single person single example problem. On the AR database, the face recognition rate based on expression and illumination datasets excluding the affect of expressions and illumination is respectively99.3%and99.0%, on the ORL database, the face recognition rate is99.0%. The method in this study has been applied in Internet cafes’Face Recognition System.