Tracking Faces Using Pca

This document describes how to track faces using Principal Component Analysis (PCA). It involves two steps: training and real-time tracking. During training, PCA is applied to a set of images containing face coordinates to compute the average image and covariance matrix. During real-time tracking, a new image is decomposed using the eigenvectors from training to obtain coordinates for the face. Downsides include needing a separate training set for each face and sensitivity to lighting, background, and movement. Possible improvements involve implementing PCA directly instead of using OpenCV.