Locating salient facial features using image invariants

Abstract

We present a method, based on the idea of salient points, for locating corresponding features between two different faces independently of scale, orientation and position. Salient points are those which have a low probability of being mistaken with other points in the face, and therefore are more likely to be correctly located in a second face. The local image structure at each image point is described by vectors of Cartesian differential invariants computed at a range of scales. Salient points lie in low density regions of the distribution of all vectors of invariants found in an image. The vectors of invariants of salient points are used to locate similar features in a second image. Results are presented, showing that salient facial features can be relocated more reliably than arbitrary facial features.