T. Lee, S. Ranganath, and S. Sanei (Singapore)
Face recognition, PCA, ICA, FA
Since the early 1990's the trend in face recognition has been the use of statistical methods as opposed to template matching. These methods are more robust to variations in lighting, facial expression and pose. Here we attempt to give an overview of recent trends as well as the mathematical basis which is in the use of linear Gaussian mixtures. The original face data is reduced in dimensionality. Statistical methods are then used to enhance the face features so it can be classified better. When a probe image is presented, various statistical techniques are used to decide to which class the face belongs. We also consider various ways the face data is transformed.
Important Links:
Go Back
