Face Recognition with Time-Varying Expressions for Homeland Security Applications
Fig-1: All images are registered using selected control points, related to the two eyes and nose. Each pair of control points is determined using normalized cross-correlation. The images are then registered using a linear transform. The image above shows results of image registration. (a) Corresponding point selection (b) Unregistered (input image); (c) Base image; (d) Registered image. |
Fig.2 Registered data base of faces |
Fig.3: Example of images used in the unregistered (raw) vs. registered images investigation. (a) whole raw; (b) whole registered; (c) partial raw; (d) partial registered; (e) ellipse raw; (f) ellipse registered |
Fig-4: Effect of varying the amount of information in an image on the recognition accuracy |
Fig 5: Effect of varying the amount of information in an image on the recognition accuracy (no off-the-plane rotation). |