摘要 |
A method for image alignment includes acquiring a facial image of a person 28 and using a discriminative face alignment model to fit a generic facial mesh to the facial image to facilitate locating of facial features. The discriminative face alignment model may include a generative shape model component and a discriminative appearance model component. Further, the discriminative appearance model component may be trained to estimate a score function that minimizes the angle 36 between a gradient direction 32 and a vector 34 pointing toward a ground-truth 24 shape parameter. The vector 34 may designate the ideal travel direction, i.e. the vector that points directly towards the ground truth. The core of the method may be known as an Optimal Gradient Pursuit Method (OGPM). |