A 350px by 300px image contains 105,000 individual pixels. Comparing each pixel to tell whether or not two images are the same is not efficient. What if we could reduce the number of features, while still maintaining patterns and trends? What if we could perform this recognition by only comparing 25 data points?

Principal Component Analysis is a standard method of extracting features from such a set of data.

This talk will show how Principal Component Analysis and the Singular Value Decomposition can be used to extract features from images of faces. With the ultimate goal to recognize the same face across different expressions and images.