Face Morphing code - October 2013


Project for 15-463 Computational Photography at Carnegie Mellon University, Fall 2013.

This project does a face morph between different people in the class. In my case, I did a morph between me and euboweja. The method of morphing I used was pretty standard. A Delaunay triangulation was computed from the average of the 43 control points selected from both images. This triangulation was used to find the position of the morph at a particular frame with a simple linear interpolation. The pixels in each triangle were then computed with a reverse transformation from target to the original image. Linear interpolation was used to get pixel values for the warped image.

I also computed the mean face by first computing the mean position of all the control points on all the class faces. I then morphed each face to the location of that mean mesh, and averaged the values.

Project Results Website