Automatic Human Age Estimation Of Facial Images Based On Feature Integration

Auto age estimation based on facial images and face verification across age aretwo very significant new hotspots, as well as challenging study and importantresearch aspect in the face recognition field. We proposed Super-resolutionreconstruction algorithm and feature integration method for age estimation, and alsoused Active Appearance Model as a discriminative method for face verificationacross age progression. The major contents and relative achievements are listed asfollow:1. A method of age estimation based on Super-resolution reconstruction algorithmhave been proposedA Super-Resolution Reconstruction algorithm was proposed to implement the ageestimation of facial images, which cut the facial image into small pieces. Then afterbuilding high resolution images by using Super-Resolution Reconstruction algorithm,the RBF neural networks was used to training and testing these high resolutionimages. At last, the classifier ensemble with genetic algorithm was used toestimating age information.2. An effect of image magnification method on age estimation has been discussedLow resolution facial image magnification and normalization is a key step inautomatic age estimation. For the problem whether traditional image magnificationmethod to some extent, affect the age estimation performance, we designed effectivecontrast experiments. In this paper according to different methods of imageprocessing and feature extraction, the experiment was divided into four groups. Then,we compare super-resolution reconstruction method with general magnificationmethods. We have conducted experiments on a large scale age databases. Theexperimental results better explain the problem proposed.3. An method of age estimation based on global and local feature integration hasbeen proposedWe proposed a novel approach combine the global and local facial features in parallel manner to implement the age estimation. Then after extracting global andlocal features, these features are integrated for fine age estimation. In the proposedmethod, global and local features are extracted by Discrete Fourier Transform (DFT)and Gabor Wavelets Transform (GWT) respectively. Radial Basic Function (RBF)network has been adopted as predictor as well.4. An AAM (Active Appearance Model) as a discriminative approach has beenproposed for face verification across age progressionWe rebuild the shape and texture model of AAM and use it for extracting agefeatures of a new facial image, by which satisfactory results of face verificationacross age progression have been achieved. Similarly, SVM (Support VectorMachine) has been used as classifier to perform face verification on FGNET andMorph.