Perspective distortion modeling, learning and compensation

We describe a method to model perspective distortion as a one-parameter family of warping functions. This can be used to mitigate its effects on face recognition, or synthesis to manipulate the perceived characteristics of a face. The warps are learned from a novel dataset and, by comparing one-parameter families of images, instead of images themselves, we show the effects on face recognition, which are most significant when small focal lengths are used. Additional applications are presented to image editing, videoconference, and multi-view validation of recognition systems.