Project 4: Face Morphing

By: Nikhil Sathe

Computing the "Mid-way" Face

The goal for this part was to compute the mid-way face of two images, A and B. The process involved computing the average shape of the two images, warping the two source images into that mid-way face, and lastly, averaging the colors from the two source images. Computing the average shape required triangulation using the Delanauy triangles. Warping required computation of the affine transformation matrix that mapped each source point onto a point in the mid-way image. Shown below is an example of a mid-way face.

Image A	Image B	Mid-way Face
Draymond Green	Stephen Curry	Combination

The Morph Sequence

The morph sequence involved changing the warp and cross-dissolve fractions over some number of frames. Shown below are two examples of morphing sequences. The first sequence morphs Draymond Green into Stephen Curry. The second sequence morphs me into one of my friends.

Sequence 1	Sequence 2

Mean Face

To calculate the mean face of a population, I used the Danish dataset images and their respective points. The results are shown below.

Results

General Mean Face	Male Mean Face	Female Mean Face

Here is a morph from one of the faces from the population to the average face.

Here is an example of my face warped into the average geometry as well as the average face warped into my geometry.

My Face	Average Face
Average Face to My Geometry	My Face to Average Geometry

Caricatures

To create a caricature, I took the difference between my face and the average face, multiplied that by a factor, and then added the result back to my image points to emphasized the unique features of my face.

Original	Caricature

Bells and Whistles

(Theme) Here is a morph video of the Warrior's Stephen Curry, Klay Thompson, Kevin Durant, and Draymond Green.

Warriors

(Student Group Morph) Here is a morph video of students from the 194 class.