| Title | The Study on Face Recognition Method and Illumination Invariant |
| Abstract | Nowadays, Face recognition technology(FRT) is one of the most active research topics in areas of pattern recognition and artificial intelligence both in china or other countries, and it has been a leading subject full of vigor and hope in science and technology. Face recognition technology belongs to a kind of Non-Touching identification technology, compared to some Touching identification technologies, such as fingerprint identified, iris identified, face recognition has many advantages. Since the 1990′s, due to the application requirements and the breakthrough in related theories, face recognition has caused widely concerns by related department and academic circles, become a hot point of research area at its peak.Although face recognition system is a biometric technology which possesses great developable potential, it has not been applied in reality on a large scale. On the one hand, human face image appearance has potentially very large intra-subject variations and diversity, on the other hand, there are many unknown factors lowering and degenerating the quality of the face images in the process of the image capture process, such as light source, objects in the environment and so on. All of these problems have let the real-time face recognition task very difficult. How to recognize the individual identity from face images correctly and efficiently to satisfy real time requirement calls for immediate solution in face recognition technology. In addition, the face images include huge amounts of information, even hundreds of digital pixel data inputed, how to extract most effective feature information is also one of the most challenging technologies.The chief contents in this paper include the following items, on account of low recognition accuracy of the face recognition algorithms based on SVD, according to analysis on this problem, three improved class estimated basis space singular value de-composition methods are proposed in the paper. In the feature extraction process, a new face recognition method based on CSVD and non negative matrix factorization(NMF) is presented. How to solve the random initialization problem in NMF algorithm always all is the research hot spot and the difficulty, a new method SVD-Based which reduces the iteration times and training time of NMF effectively is presented to initialize W and H. At the same time, SVD-Based could be combined with many current improved NMF algorithms commendably. These combinations could improve the recognition accuracy and meanwhile, reduce the iteration times and training time. For experimental database, the better recognition performance can be obtained.Face recognition technology has been developed rapidly recent years, but illumination problem which is one of the major factors of hindering face recognition technology applied into practical application has not got a good solution so far. General face detection and recognition algorithms are based on an assumption which is the face images are achived below proper and even illumination, otherwise illumination in life tend to be uneven. Illumination normalization which makes face detection and recognition accuracy a sharp drop is always an important and difficult task in the field of face detection and recognition. This paper analyzes the features of wavelet transform, make use of wavelet-based denoising model to solve illumination problem in face recognition. The experimental results prove that the strategy are valid and could be applied into real time face recognition system. |
| Category | Internet |
| Keywords | Face recognition, Illumination, NMF, SVD, wavelet, |
| FileType | |
| Pages | 185 |
| Price | US$60.00 |
| Buy Now | |
| Download | |
| Contact |
E-Mail:itpaper@hotmail.com TEL:1-888-786-998A |
| FAQ |
How to get this paper's electronic documents? 1, Click the "Buy Now" button to complete the online payment 2, Download the paper's electronic document from the successful payment return page/Or the system will send this paper's electronic document to your E-Mail within 24 hours |
| Favorite | ADD TO FAVORITE |
| Version | zh-cn |
The Study on Face Recognition Method and Illumination Invariant
Billy HU, Ph.D
E-Mail:itpaper@hotmail.com
TEL:1-888-786-998A
