CS194-26: CS194-26: Image Manipulation and Computational Photography

Project 3 – Face Morphing

Dimitrios Vlachogiannis (St. ID: 3034311700)

Defining Correspondences

Using ginput we select points from the desired images according to the following order:

We then define the triangulation on the selected points. In our case, the employed triangulation method is the Delaunay triangulation as it maximizes the smallest angles of the triangles producing more evenly sized triangles.

The example images that we used to apply the morphing framework are the ones below of the best duo in Los Angeles:

 

Original Lebron                                                                                    Triangulated Lebron

Original AD                                                                                          Triangulated AD

 

Computing the "Mid-way Face"

As an intermediate step before the morphing we were asked to produce the mid-way face by computing the average key points correspondences, wrapping both images to the average shape and mixing them together:

The Morph Sequence

As a generalization to the process of the mid-way face, 60 evenly spaced intervals were generated from 0 to 1 to regulate the transition from Lebron to AD. The generated video is on YouTube under this link:

 https://youtu.be/kxV9zBRmj34

The "Mean face" of a population

For this part, the IMM Face Database was utilized and particularly the mean female face was calculated:

 

Caricatures: Extrapolating from the mean

As a caricature, image of a man was extrapolated from the population mean calculated in the previous step:

Original Image of man                                     Man triangulated                                            Man wrapped to average female shape                             Caricature

 

Bells and Whistles

For bells and whistles, all the happy images were used as base population to turn a sad man smiling:

Original Image of man                                              Man triangulated                                                      Man wrapped to average smiling shape                  Man smiling