Face Morphing

Aaron Johnson, Computational Photography Fall 2007 Proj3

The third project for this class involves morphing faces into other faces. To do this we first compute a triangular tessellation of the face's geometry, find a (weighted) average of the tessellations, then streach the pixels from the two inputs into this average shape, and blend their values. Sounds complicated but it isnt really. Lets see some simple results. First me, then the morph, then wwedler:




Pretty cool, isnt it! You can even animate the transformation:

Those animations are kinda large though so from now on i'm just going to show you one frame and you can click on it to see the animation.

Here is the grid overlayed on our images:



One thing I did was that in additonal to all user specified control points I added points at the corners and midway along each side. This resulted in decent transformations of the background. Here is me transforming into some other classmates, and their origonal picture on the right. The first of these was my contribution to the combined show, so the animation is longer and made with twice as many control points, including some to define the part in the hair, shirtline, sholders, etc:



Here is me selecting the extra control points for the first of those. Note that points 7-9 were later dropped because they did not have the desired effect of growing the tuft of hair out of my hair:


Transforming faces into other faces gets boring quickly. So I thought I would try transforming faces into gorillas :



Not bad, how about some other people (morph on left, orig on right, duh):



And here is my face just warped into the shape of a gorilla's:


What if we could merge all the faces into one uber-face. Amazingly that's what I did next:

I would totally believe that he/she was in my class...
With this we can go on and figure out many interesting things, such as what would I look like if my bone structue was that of the average student, but skin tone stayed the same (1 geometric distortion, 0 color)

Apparently I am less asian than the class...How about the other way around, my face, average color? (0,1)

Hm, I guess my cheeks are droopier. What if we exaggerated my features?(-1,0)

My smirk is growing! How about factoring in some color differences?(-1,-.5)

Suddenly my hair is darker, shirt is lighter, and my beard has grown some. Even more?(-1,-1)

Even more but now I look badly burned (i guess I was tanner than average?) How about even more exaggerated features?(-2,-.5)

Well every algorithm hits it's limits at some point...Thinking back to when I turned myself into a girl earlier in this assignment, I decieded to over-compensate. Here is a charicature of me with respect to amichals, (i.e. over-male) (-.5,0):

Even more?(-1,0)

Wow those triangles are broken. What about just me, but with her face structure (i.e. female)?(1,0)

That was fun. Why would I want to leave my classmates out of this fun? (all -1,-.5)