Face Recognition Across Pose

In many face recognition tasks the pose of the probe and gallery images are different. In other cases multiple gallery or probe images may be available, each captured from a different pose. We have developed a face recognition algorithm that is able to handle large variations in pose using a “patch-based” rather than “holistic” representations. The algorithm operates by estimating the “eigen light-field” of the subject’s head from the input gallery or probe images. Matching between the probe and gallery is then performed using the eigen light-fields.

Our algorithm drammatically out-performs, FaceIT, a commercially available system from Visionics Corporations. A brief summary of our results on the PIE Database
are: