In this project I implemented face morphing. This is done by first specifying corresponding points in a series of images, and then creating a vector field of warp vectors which morphs one image to the other.
This vector field is calculated in a similar technique to poisson blending, where the field is specified at each user supplied point, and the field at any other point is the everage of its neighbors. I determined two vector fields, one for x displacement and one for y displacement.
One subtlety of the implementation is the use of inverse warping; sampling each pixel in the target image from some location in the source, rather than projecting each source pixel to the target image. This means that the vector field is measuring the negative of the amount which the image is shifted along a given axis.
I created two videos of image morphing, one of the class photos and the other of the classic computers throughout the cit. The movie files are large, but from a cs dept machine you can see them at /home/chonger/movies/computer.avi and /home/chonger/movies/example.avi.
