AGE INVARIANT FACE RECOGNATION

Abstract

ABSTRACT Age invariance face recognition( improved techniques ) is an emerging field of engineering in signal processing. The face recognition is a technique prior to the age invariance face recognition technique. To achieve the face recognition in the computer we have to put some image in the computer and by some algorithm it recognize the whether the image has any human face or not however. As it is emerged from the extensive research in the different discipline in the signal processing and image acquisition engineering. There are various steps before the actual achievement of the age invariance face recognition and acquisition devices. First of all the acquisition devices there are various acquisition devises which are video camera, laser, thermal, acoustic, UV, act. these sensors gives the input to the algorithm which uses the input to find the proper result. Second thing is the algorithm which recognizes the human face from the given data image there are several approaches to the face recognition like posing, illumination, occlusion, edges, contour, color, scale, this are the factors which are used to find the face of the human being. In the image processing the first task is find the pixels and then the mask then the finding on the lines and edges and the contours then the various shapes which in turn will help the computer find the algorithm which finds or detects the shape of the object and then the computer find whether the shape resembles the human face or not the other features are detected. soon the human face is detected the next task is to find the algorithm which classifies that face with data in the data base. there are several techniques to find the face recognition. Face recognition system is a system which rely on the data stored in the computer storage images of the human subject are taken and stored and at the time of the actual face recognition the image are compared with the data images and the results are compared with different standard data set for confirmation. There are many techniques which currently being used to recognize face and are employed in a very wide applications such as identification in banking data, security areas where authenticity of a person is critical and must provide real time identification, in surveillance of large area and many more applications but one that the face recognition algorithm get stuck on is that identifying the person who has been aged and has had a different facial features now to find the algorithm which can recognize a person regardless of its age. To accomplish this there have been many technique have been proposed to tackle that problem in analytical way. As the variation in the face in very unpredictable and terribly complex as we know intuitively that every person ages differently and one analysis a cannot be used to determine the ageing of other person. one of the technique is using SELF PCA based method in which self Eigen and Periocular Region is used. Other method is MULTIVIEW DISCRIMINATIVE LEARNING FOR AGE INVARIANT FACE RECOGNITION in this method it is taken in the account that the local features are more robust to age variations in this method there are three local descriptors are taken in account these are SIFT (scale invariant feature transform), LBP (local binary patterns), GOP(gradient orientation pyramid) and the results are tested on the FG-NET face aging dataset for efficiency. there are another method called A DISCRIMINATIVE MODEL FOR AGE INVARIANT FACE RECOGNITION in this SIFT,MLBP(multi-scale local binary pattern) serve as a local descriptor and to avoid over fitting it uses MFDA ( multi feature discriminant analysis), the other method which is based on PERIOCULAR BIOMETRICS in this WLBP ( Walsh- Hadamard transform encoded local binary patterns) and UDP (unsupervised discriminant projection) collectively. so far we are done with the feature extraction now the next step is the dimension reduction the reducing the dimension is very important because it reduced the redundant data in the procedure and we have few technique to do it first is PCA ( principle component analysis ) there is also another method called NONLINEAR TOPOLOGICAL COMPONENT ANALYSIS in this method KRBF is used to reduce the data and alpha shaped constructor is used extraction and various classifiers are used to classify the image. In this thesis, I have proposed method in which the reduction of data is done by the combination of the KFBR and MFDA, features extraction is done by the combination of MLBP,SIFT and LBP and face matching is done by multiple LDA based classifier and maximum a-posteriori probability Gaussian mixture by using adaptive boosting algorithm.