Age Estimation Of Facial Images Based On Gabor And Histogram Of LBP

Human face conveys much important perceptible information related to individual trails, such as gender, skin color, race, age. Study on the human facial images has a significant meaning and value. In recent years, age estimation of facial images is also intensive exploration, and some greater progress has been made. However, the accuracy of age estimation should be improved, which is constrained by the facial expression, posture, accessories, light and other factors. On the basis of the studies at home and abroad, the facial feature extraction, age estimation and feature dimension reduction are researched and the experiments have achieved certain effects. The work of this article is described as follows:(1) Localization of facial image. Because there are different sizes and areas in the database of facial images, all of images in the database are processed by the method of face detection based on Haar and AdaBoost algorithm into a new experimental image database, each facial image in which has the same sizes. It effectively reduces the interference from the non-face and improves the accuracy.(2) Texture feature extraction of facial images. In recent years in the field of image processing, face recognition and so on, Local Binary Pattern (LBP) and Gabor wavelet methods are commonly used. Some experiments are organized in the article and the feature extraction method proposed is conformed.(3) Dimension reduction of feature space. Principal Component Analysis (PCA) and histogram statistics of image blocks that is broad applied in the face recognition field are used to get the original feature dimension reduction. And the time consuming and age estimation accuracy are analyzed and compared.(4) Age estimation of facial images. Research in the application of Support Vector Machine (SVM) in the age regression, Gauss Radial Basis Function (RBF) based on Support Vector machine Regression (SVR) is used to train and test the face aging database, by using the LOPO strategy.The experiments in this article conform that the different texture feature proposed is effective to age estimation of facial images. And the experimental results show that a high accuracy and appropriate time consuming are achieved.