Face Recognition Experiment Using 2DPCA and 2DLDA on ORL Face Database

We conducted comprehensive face recognition experiments using 2DPCA (Two-Dimensional Principal Component Analysis) and 2DLDA (Two-Dimensional Linear Discriminant Analysis) algorithms on the ORL face database. The implementation includes detailed code annotations covering key steps such as data preprocessing, feature extraction using matrix-based operations, and classification procedures. For 2DPCA, the code demonstrates how to compute covariance matrices directly from 2D image matrices without vectorization, while 2DLDA implementation shows between-class and within-class scatter matrix calculations for better class separation. This experiment serves as an excellent practical guide for beginners to understand the fundamental differences between PCA (unsupervised dimensionality reduction) and LDA (supervised classification) algorithms. Through hands-on implementation, learners can grasp how these algorithms transform facial images into feature subspaces and perform recognition using distance metrics like Euclidean or cosine similarity. Our findings indicate that both 2DPCA and 2DLDA achieve high recognition accuracy on the ORL dataset, demonstrating their strong potential for future face recognition research and real-world computer vision applications.