CS 194: Computational Photography, Fall 2017

CS 194 Proj 4: Face Morphing

Anshul Tibrewal, CS194-aam



Overview

Part I: Face morphing

Algorithm

In my algorithm I find 14 correspondance points through which the images are going to be trinagulated form one to another. I use a Delaunay triangulation of the mid-distance points for the two images where the equation is:

middistance = t*shape_im1 + (1-t)*shape_im2 /p>

I do a similar thing for color between the two images. I generate 45 images for different fractational time stamps ranging between 0 and 1 and put all of them together as a gif file.

I morphed my friend's face with mine and the gif is displayed below:


Face Morph with my friend Colin

Part 2: Mean faces of the frontal features dataset

I downloaded the dataset here: http://fei.edu.br/~cet/facedatabase.html.

Then I computed the average image of this dataset and I display it below.

Avergae Face of the Frontal Feature Population