To find the mid-way face, or average of two faces, I decide on corresponding points between the two faces. Then, we find a Delaunay triagulation on the average of the two sets of points. For each triangle in the triangulation, we take find the affine transformation for each set of source points to the warped points, continuously morphing the images. Then we cross dissolve for the color. Here is the mid-way face of my face and my friend, Anstonia.
We can create a morphing sequence from one face to another using the same technique as the mid-way face, except with varying the input weights for the warp and cross dissolve based on our current step. Here I morph my face to my friend's.
Here I found the average male smiling face of the Danes. I computed the average shape from the annotated data provided, then generated a morph sequence from each person to the average shape. Then taking each image that was fully warped to the average shape, I take the average to find the average smiling face.
Male 1
Male 6
Male 7
Male 16
Male 28
Male 13
Male 34
This is the average male smiling face, along with that face warped to my face shaped, and my face warped to the average.
Original average smiling and me
Me to average shape, Average to my shape
Warping a face away from the average geometry will produce a caricature. Here I warped my face away from the average smiling male Dane to produce a caricature, exaggerating my features.
Here I take the image of the average Han Chinese male to change my appearance.
Originals
Shape, Appearance, Both