### Defining Correspondences

We first defined pairs of corresponding points on image A and image B. Here we chose 46 points which covered the outline of the head, the facial features, and the 4 corners of the image. Then, we take the average of these two points to define our triangulation.
Here's an example of triangulation which captures most facial features of the image.

### Computing the Mid-Way Face

To compute the mid-way face, we took the average points from the previous part and created a Delaunay triangulation. Then, we computer the affine transformation matrix from the average triangle to the corresponding triangles of image A and image B.
Here we have a woman and a man, and computed the mid-way face.

### The Morph Sequence

To produce a sequence, we warped using the previous steps on various fractions which will let us emphasize image A more in the beginning and gradually focus solely on image B.
Here are our results:

### The "Mean Face" of a population

We picked images of bearded men from the Danes dataset. We computed the average face of these bearded men by morphing each of these faces into the average face.
Here is an example of a bearded man from our dataset

After morphing each face and computing the average, this is average face we ended up with:

### Caricatures: Extrapolating from the mean

With the average bearded man, we can create caricatures!
Here is a picture of me:

Here is again the average bearded man:

Here is the morphed shape:

Here is the caricature of me blended with an average bearded man:

### Bells and Whistles

I changed the ethnicity of my face by using the average Korean woman and morphing it with my face. Here are the results:

Average face of Korean Woman

Me

Morphed shape

Changed ethnicity