Face detection based on eigenfaces and legendre moments

This paper presents a new approach for face detection based on eigenfaces/principal component analysis (PCA) and Legendre moments (LM). PCA and Legendre moments are two different methods used for detecting patterns in images. We present a hybrid system for face detection which combines the eigen weights calculated by PCA and Legendre moments calculated by Legendre polynomial together. These combined weights and moments are then used to train a support vector machine (SVM) for classification. Our approach performs better when compared with the individual approaches. With 300 face images collected from ORL database and 200 non-face images, it produces 96% accuracy (verified by 10-fold cross validation), which is better when compared with the individual approaches and the previous works such as.