Study Of MGA Based Face Recognition

Face recognition technology, which is a biometric identification technology, hasbeen widely used in practical applications. However, there are still many factors thataffect the face recognition results, and the problem of shadow developed by unevenillumination is a major factor. This paper makes full use of the multi-scale,multi-directional and localized features of a new generation of multi-scale geometricanalysis tools - Contourlet transform, improves the result of the shadow compensationof the face images under uneven illumination and extracts the invariants of the shadowarea of the face images under uneven light conditions effectively.This paper details the development and construction methods of the discreteContourlet transform, and leads to a continuous Contourlet transform, and then we dothe simulation experiments of the Contourlet transform. It has a multi-scale,multi-directional and localized characteristics. We realized the 2D image shadowcompensation algorithm and the Contourlet based shadow compensation algorithm, butthe two kinds of algorithms exist their own strengths and weaknesses, so we willcombine the advantages of both, and we can extract the detailed information of theshadow part and retain the non-shaded area information effectively. This combinedmethod improved the image quality effectively and for the face recognition provides agood shadow compensation method of uneven illumination for face recognition. Finally,we introduce the total variation model, combine the advantages of total variation modelwith the advantages of Contourlet transform, present a method of Contourlet transformbased TV model to extract illumination invariant. This method can remove theoscillation phenomenon effectively, and can eliminate local constant produced by TVmodel to a certain extent, so there is a higher recognition rate.