Facial Feature Rearch Based On Wavelet Transform

Face recognition is an identity recognition technology with history more than 70 years,continuing mature of technology promotes technology out of the laboratory into the production and life practice.Successful face recognition lies in an effective process on High-dimensional data feature,influences of different light conditions and other factors on Recognition performance,and ask for a high recognition rate and time efficiency.Currently,the face recognition technology has been widely used in National,International Safety Management and Commercial and teaching activities such as security,monitor,Multimedia Management,HCI and Authentication,and so on.Wavelet transform is provided with Multi-resolution analysis feature and fast efficient Mallat algorithm;Uncorrelated Linear Discriminant Analysis can extract Statistics irrelevant features.Combining the advantages of wavelet transform and ULDA methods.Wavelet transform can solve the small sample size problem of ULDA.ULDA can extract face images' Statistics irrelevant features.low-frequency and high-frequency both extract their ULDA feature.Finally,using four sub-graph'Fusion features to classify.To prove the algorithm's effectiveness on improve recognition performance by simulation experiment.To study Illumination invariant feature extraction algorithm based on Wavelet transform. On this basis, using histogram equalization to reinforcing illumination invariant features‘ contrast.Using comparing experiment to find out the best Wavelet,Training set and Test Set's Parameter combination,which can make the Recognition performance at their best.Experiments on light face database show that improve algorithm has a better recognition rate and Time efficiency.Proposed Illumination invariant feature extraction algorithm by fusioning Wavelet Low Frequency information.In order to take full advantage of Wavelet Low frequency information,Using Gaussian smoothing filter to re-extract the Minutiaeinformation from Wavelet low frequency,as a useful complement illumination to invariant features.Fusion Algorithm has richer Characterizations to represent a image affected by light.