CS194-26 Project 3 Face Morphing

Defining Correspondence

To define keypoints for warping, I used ginput to collect 47 points on the image, including 4 points in the corners.

Image 1 with correspondences
Image 2 with correspondences

Mid-way Face

We can compute the 'mid-way face' of two images by weighing them differently each time.

A picture of George Clooney
Mid-way Face
A picture of Zlatan Ibrahimović

Morph Sequence

Now, by producing multiple 'mid-way faces' by increasing the warp fraction and weighing the pixels we can create a MORPH

Morph Sequence of George Clooney to Brad Pitt.

The "Mean face" of a population

By calculating the average shape of a population's faces, and warping each of the images to the average shape, we get to the 'mean face'.

I used the FEI Face Database, and used images of 20 females to get an average female face shape. Here are examples of warped images:

Female_Image 29a
Warped Female_Image 29a
#21 Female_Image 43a
Warped Female_Image 43a

Now, by adding up all the warped images as above, we can get the mean faces.

Average Face
Mean Face

Warping Me(A Generic Man Image) to Mean and Mean to Me(A Generic Man Image)

Original Me
Me warped to Mean Image
Mean Image warped to Me

Caricatures: Extrapolating from the mean

Extrapolating from the mean female face, we try to exagerate certain features to create caricatures:

With Alpha=-0.5
Original Image
Me with alpha=0.5
Me with alpha=0.8

Bells and Whistles

Changing Gender

I morphed a man's face to the mean female face to see how a gender change would look like for him. This is the result: a) His jawline has become much software and the shape of the face is much narrower like that of females. b) His hair on the sides is warped to become slightly longer. c) Because of his glasses it was harder to warp the area around his eyes

Mean Female Face
Mean Image warped to Me