Project 3: Face Morphing

Evan Huang

Part 1: Defining Correspondences

Evan	George
Evan triangulated	George triangulated

I preprocessed the images by cropping/resizing them to the same shape and using this website to make the backgrounds white. I then used the online correspondence tool to label the images. I used the Delaunay triangulation on the averaged points.

Part 2: Computing the Midway Face

Evan

George

Evan/George Midway

I computed the midway face by computing the affine transformation matrix between each corresponding triangle. I then used inverse warping and linear interpolation to determine the coloration of each triangle.

Part 3: The Morph Sequence

I generated a morph sequence gif by finding different weighted averages of each image, determined by warp and cross-dissolve fractions. The following gif has 46 frames.

Part 4: The Mean Face of a Population

I retrieved images from the FEI Dataset. Using the given points, I triangulated the images based on the average points. I then morphed a set of 35 neutral-faced (not smiling) images to the average shape.

Original Image	Morphed Image

Taking the average of all 35 images gave the following average image:

I then morphed my face into the average geometry and vice versa:

My face to average FEI dataset geometry

Average FEI dataset face to my geometry

Part 5: Caricatures; Extrapolating from the Mean

I created a caricature of myself by morphing my face to the average geometry from the FEI dataset but setting the warp fraction to 1.5:

B&W: Changing my Age

I changed my age by morphing my face into a picture of an elderly Asian man. I followed the same process as above, labeling the points using the online tool and morphing the images.

Original images:

Morphed images:

Shape only	Appearance only	Both Shape and Appearance