An enhanced face recognition with modular locally discriminating projection

LDP is a supervised feature extraction algorithm hence it considers both class and label information for classification. A new face recognition algorithm named Modular Locally Discriminating Projection (MLDP) is presented in this paper. In the proposed method, initially the training and test face images are subdivided into smaller sub face images and then LDP is applied to each sub face images. Within this sub face images some of the local features do not vary largely corresponding to pose, lighting and facial expression of individual face images. The proposed MLDP captures most of the similarity features against pose, lighting and facial expression. This improves the classification accuracy. The experimental results on the ORL face database suggest that the proposed modular LDP has better recognition rates than Modular PCA and other conventional feature extraction methods.