Contourlet and nearest feature line based feature extraction approach for one prototype sample problem

In this paper, a novel algorithm for face recognition with one sample per person is proposed. The proposed algorithm is based on contourlet transformation. For simple prototype sample problem, many discriminant learning methods can not work. Because for most discriminant learning methods, the within class scatter of the prototype samples are very important. However, simple prototype sample problem does not have within class scatter. To enhance the representative capability of the prototype samples set, some new samples are generated using contourlet transformation. Multiple prototype samples for each class are constructed through the decomposition and reconstruction of original training images by contourlet transformation. Thus neighborhood discriminant nearest feature line analysis can be performed on the new database. The experimental results demonstrate the efficiency of the proposed algorithm.