Face recognition based on data field

Principal component analysis(PCA) is one of the important methods for extracting the main features of face images. However, it mainly considers the values of the pixels in the images while ignoring the relations between one pixel and other pixels around it. In fact, the pixels of the face images are not independent of each other. Based on this, we propose a new method for extracting the features of face images based on the datafield method which considers the relations between pixels. Furthermore, we combine the data field with PCA(dfPCA) to attract more accurate features from face images. In the experiments, we first analyze the contours of face images with datafiled method. Then, we apply the dfPCA methods on two real world datasets, ORL and yale, for feature extraction respectively. The results show that the recognition rate based on the dfPCA method is better than the PCA method, which demonstrates the feasibility of datafield method for extracting facial features.