Morphing into a Another Student

To morph between two faces, I start by specifying corresponding control points on each image by hand. Then, the average of the control points is taken, and the Delaunay triangulation of this set is calculated. Every pixel on the output image is then converted to barycentric coordinates in relation to the triangle that contains it. The reverse transform is done using the corresponding control points in the original images to get the corresponding color. Then, the weighted average of these colors is taken.

Morphing into the Class Average

Taking the average of a set of faces is a similar process. Every image needs to have corresponding control points, which are then all averaged together to calculate the Delaunay triangulation. Every image has its geometry morphed to match the average, but the original color is kept. All of these results are then averaged together.