In this project, we explore face morphing, population means, and caricature generation through image warping and cross-dissolve techniques.
In this section, we manually labeled corresponding points on Elon Musk and Christian Bale's images to define key features like eyes, mouth, and chin. We then computed a Delaunay triangulation for the morphing process.
We computed the average shape between the two faces, warped both faces into this shape, and averaged the colors to produce the mid-way face.
Using our defined correspondences and the morphing function, we generated 45 frames of smooth transformation from Elon Musk's image to Christian Bale's image. Below is a GIF of the morph sequence.
Some other examples are also provided below
We computed the average face shape from FEI face dataset of annotated faces, warped each face to the average shape, and then computed the mean face of the population.
Here is my face warped into the average geometry, and the average face warped into my geometry
By extrapolating from the population mean, I created a caricature of my face. Below are the results of this exaggeration.
Finally we can get creative with the images and construct some of the functionalities of liquify filter from photoshop. Here we have the picture of Daniel Radcliffe twirled and shrunk