In this project, I explored some of the interesting image editing techniques that can be done with a dataset of similar images with labeled feature point--namely my classmates' faces! The starting point each of these techniques is the local-warp operation. In this operation, we first find a triangulation of a sparse set of feature points. Next, we move each of the feature points in an image to a new position, using the triangulation to define an affine warp for the in-between pixels.

Face Morphing

With the ability to locally-warp images, we can morph one face into another. We simply start by warping each face into an in-between shape, and blend the warped images together.

Averaging Faces

Not only can we warp two faces together, we can also morph any number of faces together to find an average face.

Subpopulations

If we look at subsets of the class that share a characteristic, we can take a local average instead of a global average. This lets us answer questions like ''what does the average beard look like''?

From left to right: the average Computational Photography student this year, my face warped to the average shape, and the average student warped to my face's shape.

Here are some average faces over different subpopulations of the class:

Bearded students

Non-bearded students

Students of eastern descent

Students of western descent

With these local averages, we can now change characteristics of other faces by adding in some of the averages.

Beard addition	Beard reduction
Easternization	Westernization