FisherFaces for Face Matching

FisherFaces for Face Matching is a cool tool that lets you create and tweak faces in a 3D space! It’s designed to help with face recognition, which means it can identify faces even when the lighting changes or if someone makes a funny expression.

How Does It Work?

This software uses a special algorithm that looks at each pixel in an image like it's a point in a huge space. The neat part is that images of the same face, even if the light changes but the pose stays the same, actually fit into a specific 3D area of this high-dimensional image space. This only works well if we assume the face behaves like a Lambertian surface – basically, no shadows!

Dealing with Shadows

But let’s be real: our faces aren’t perfect Lambertian surfaces and they do cast shadows. So, instead of trying to map all those shadows, FisherFaces smartly projects images into that special subspace while ignoring areas with big lighting differences. This is done through something called Fisher's Linear Discriminant, making sure different classes are clearly separated, even when there's lots of change in light and expressions.

The Power of Fisherfaces vs Eigenfaces

You might have heard about another method called Eigenface. It's similar because it also projects images into lower dimensions. However, tests on famous databases like Harvard and Yale show that FisherFaces often has lower error rates than Eigenfaces. That’s pretty impressive!

Key Features

If you're interested in diving deeper into what FisherFaces offers, check out its main features:

• Face Recognition: Great at identifying faces under different conditions.


• Illumination Invariance: Handles various lighting scenarios well.


• PCA: Uses principal components analysis effectively.


• Download FisherFaces here!

Requirements

You will need:

• Matlab Image Processing Toolbox