Face Morphing

Kyle Hua, CS194-26-AGX

Defining Correspondences

I chose 40 points on the faces, with the corner points for a total of 44 points. Then I used Delaunay to create triangles based on the points. I used pictures of me and Frederick Kim.

Computing the Midway Face

To compute the midway, I computed the average shape between the points. Then warped the pixels by computing affine functions to warp the pixels. Then I computed the average of the colors.

The Morph Sequence

Computing the morph sequence used the same method as the midway, but we adjusted the weight of pixel shape and color from a range of 0 to 1. This linear combo created images from ranging from completely Fred to a 1/4 me, 3/4 Fred to half, half to completely me.

Mean Face

Using the Danes data set. I computed the mean face of the entire population.

Face warped to mean and vice versa

I took the average points of the population to map my face onto the points and mapped the mean face to the points on my face.

Caricature by Extrapolating

I created the caricature by raising the factor of other average points on my own face.

Bells & Whistles: Changing Ethnicity

Using an online average of an italian man, I merged my own face to it. Resulting in Italian Kyle.