face

I had 100 eigenvectors and reduced to 90.

alpha = 1.

adjusted data = original data matrix - mean

param = (reduced eigenvector)' * adjusted data

reduced face data = reduced eigenvector * param

top: average - (alpha * reduced face data)

middle: average

bottom: average + (alpha * reduced face data)

Things to do during spring break

1. try to reduce eigenvectors and see how much I can throw eigenvectors away without destroying morphable model.

2. try to create caricture using reduced MM.
parameter = eigenvector * particular face
model = average + parameter * eigenvector

3. think about using neural network to evaluate parameters.


Here is sample of my thesis.
Only abstruct and will modify due to taking neural network idea into this thesis.
http://student.uncw.edu/sh0110/main.pdf