Project 3

# Face Morphing

Author: Skylar Sarabia

# Part 1: Defining Correspondences

In order to define corresponing points between two images, I built a web page which allows the user to upload two images of the same size and select correspondeces. The features of the web page include:

• Export Correspondences
• Import Correspondences
• Delete Correspondence Points
• Match existing correspondeces from a labeled image to a labeled one

Try it Yourself Click to open the correspondece tool
My Correspondences, Leo's Correspondences, The trianglulation of our average shape

# Part 2: Computing the "Mid-way Face"

For interpolation I used nearest neighbor since linear took much longer. To compute the average I used the average shape of both photos, morphed both faces to the average, then took 50% of each photo and added them together.

Me Input
Average Me & Leo
Leo Input

# Part 3: The Morph Sequence

Here is a morph sequence over 30 frames which bounce loops. For every frame the shape_frac and dissolve_frac were uniformly incremented. Its the same as the previous part except instead of evenly weighting the shape and color, it is incremented over time.

Me & Leo Morph

# Part 4: The "Mean face" of a population

I used the Danes face image database for computing the mean of the population. I filtered the database to only include images of faces in the "Full frontal face, neutral expression, diffuse light" position. To compute the mean face I first took the average shape of all the faces in the population based on their points. Then morphed all the images to that shape. Then added all the morphed images up based on their weight (1/population_total).

### Geometry Morphs

Example 1
Example 2
Example 3
Example 4
Me warped to average geometry
Average warped to my geometry

# Part 5: Caricatures / Extrapolating from the mean

Here are a few different examples of caricatures extrapolating from the mean of the danes face database.

Equation: $$my\_shape + a * (my\_shape - average\_shape)$$

Given that I have smaller eyes and a larger nose than this population average, you can see that these features are exaggerated.

a=0.25
a=0.5
a=0.75
a=1
a=1.5
a=2

# Bells & Whistles: couple2baby

For the bells and whistles I combined a photo of myself, my girlfriend, and a baby to create a concept of what our offspring could look like. I did this by weighting our geometry and colors separately. I then morphed the baby into our average photo to simulate it growing up.

Me
Meg
Baby Photo
Our Baby
Our Average
Baby growing up