Face Morphing

Algorithm

The basic image morphing algorithm takes two images, each with corresponding control points, a morph fraction, and a dissolve fraction. It first computes the average points between the images and created a Delauney triangulation. It then morphs the start and end images by transforming points within each triangle from their location in the source image to their new location according to some weighted average of the start and end control points. Finally, it computes a weighted average of these two morphed images to produce the final result.

Average Face

The average face is computed by first computing an average of the control points across the entire class. Then, each photo is morphed to this average geometry and the resulting images are themselves averaged. By providing a morph weight greater than 1, this average point set can also be used to create caricatures, exaggerating the differences between an individual's control points and the class average control points.

Morph

Images

Average	Average Pushed Towards My Points
Pushed Towards Average Points	Pushed Away From Average Points

Original Images