Greedy Face Recognition by Reduced Basis Decomposition

Efficient lossy image compression and greedy face recognition through Reduced Basis Decomposition

We test the Reduced Basis Decomposition technique for simple image compression and demonstrate its efficiency. We also devise a face recognition algorithm based on Reduced Basis Decomposition and apply it to a data set of 575 facial images to show its accuracy (comparable to PCA) and superior efficiency (better than PCA-like techniques when the number of eigen faces is up to 15% of the size of training set).