Louisiana State University
Department of Electrical and Computer Engineering
EE4780 - Introduction to Computer Vision
Spring 2006

Problem Set 5

Assigned: April 24, 2006 (Monday)
Due: May 5, 2006 (Friday) (Email to the TA (zhao.ta@gmail.com) by midnight.)


Problem: (100 points)

Your goal is to find face locations in "Test.jpg" image.

Extract faces manually from the training images. 
The number of training images you want to use is up to you.
Use principal component analysis (PCA) to find eigenvectors to represent face space.
(While finding eigenvectors, MATLAB may run out of memory.. In such a case, you should downsample your images..)

Using the eigenvectors and the average image, try to locate faces in "Test.jpg."

Suggestion: Also use skin color information. 

What to return:
1) M-file that takes an image (named "Test.jpg") as input and displays it with faces marked.
2) Other files (such as a ".mat" file that has eigenvectors in it) to be used by your m-file.


* Criteria:
1) High correct detection
2) Low false detection or repetition
3) High speed


* You may do this project in two-person teams or individually.