CS 194-26 Fall 2018

Project 4: Face Morphing

Chengzong Ou <aai>

Finding Correspondces

For part 1 of this project, we must first define corresponsding points on both images in order to morph two images together. I have manually chosen 69 corresponding points on both images, using some guidelines from the triangulation from the lecture slides. I then created Delaunay triangulations for both faces. A subtle issue is that all points must be chosen in the correct order, or the calculation for the midway triangulation will fail.

The following are two images with points I've chosen on the images, and the triangulation of the images.

Figure 1: My face with points and triangulation.
Figure 2: My friend's face with points and triangulation.

Computing The Midway Face

To comptue the midway face, we first need a triangulation. Since both images will be converging to the midway face, I have chosen to do the triangulaton on the midway of the points. This is done by averaging the all points in the two faces, and computing a triangulation there. The final result is the following image:

Figure 3: The midway face.

The Morph Sequence

Figure 4: The morph sequence from me to my friend.

The "Mean Face" of a Population

I have used the data of the Danes linked from the project website to create the following average face of a Danish male.

Figure 5: The average face of a Danish Male from 30 sample images.

The folloiwng are a few faces from the data set warped to the average structure. Original images is on the top, and the warped image is on the bottom:

I have also warped my face into the average geometry, and warped the average face into my geometry:

Figure 6: My face warped to the average geometry.
Figure 7: The average face warped to my facial geometry.


I have also applied a negative warp factor to get the following effect on my face.

Figure 7: My face with a negative warp factor using the average face.

Bells and Whistles: Changing Friend's Gender and Ethnicity

I applied the average Dane male transformation on to my female friend's face.

Shape Applied
Dane Applied to Her Facial Structure
Both Appearance and Structure