Curvelet feature extraction for face recognition and facial expression recognition

To overcome the weakness of the wavelet analysis which is unable to extract curve features of the face image, this paper applies a new multiscale geometric analysis tool - curvelet transform, for facial processing and feature extraction. A new approach based on curvelet transform and subband weighted fusion algorithm is proposed for face recognition. A novel method based on curvelet transform and support vector machine (SVM) is designed to recognize facial expression. Face recognition experiments with Yale database and facial expression recognition experiments with JAFFE database are carried out. The correct recognition rates are 93.33% and 94.73% respectively in “combined curvelet + SVM face recognition” and “curvelet + SVM facial expression recognition”. These results show the advantages of curvelet transform in facial processing and feature extraction.