Midway Face between 2 Faces

Given 2 images A and B, we can compute a midway face between them. We first define some distinguishable points on both images to use as correspondence points. Using these points, we compute a mean and Delaunay triangulation with the mean. We can use triangulation to find the appropriate affine transformation on each of the images one triangle at a time to the mean. we then average the colors together.

Computing the Morph Sequence

This part takes a sequence of "midway faces", with the only difference being the weight put on each image. We use a warp factor and dissolve factor to compute a weighted average when warping and cross-dissolving. When putting these images in a sequence, we can see the morph sequence going between the two images in 45 steps.

Average Face

Next, find the average face, not just between two peope but a whole population. We again use (preset) points to create a triangulation where each image's points are weighted equally. Similar to the previous sections, each image is warped and cross-dissolved into an average.

Average female Danish face

original face

warped face

original face

warped face

original face

warped face

Midway face of me and the Average Danish Woman

My face warped to the average Danish Woman

The average Danish woman warped to my face

Caricature

To exaggerate the more unique features of a face, we can take the difference between the face and an average face. We then add some factor of that difference back to the original face to result in a caricature.