Face Recognition Using Eigenfaces K RAMNATH BITS PILANI

Face Recognition Using Eigenfaces K. RAMNATH BITS - PILANI

Outline • • • Introduction Why face recognition ? Eigen-faces (overview) Project Evolution Future scope

Introduction This is a face recognition system, based on EIGENFACES method. The method is both intuitive, simple to express in mathematical terms, and very flexible.

Why Face Recognition • An interesting field • Lots of applications: - Security - Intelligent Computer-User interaction - Available resources

EIGEN FACES Any face in a graphic image format can be viewed as a vector by concatenating the rows of the image one after another

The Eigenface Method Eigenfaces is to a set of basis functions that if linearly combined could represent any face.

EIGENFACES Eigenfaces are used for the recognition and detection of (among other things) human faces.

Eigenfaces • These photographs comprise a training set, and it is this training set that the eigenfaces are extracted from. • The photographs in the training set are mapped to another set of fifteen images which are the eigenfaces. • The photographs in the training set is one domain, and the fifteen eigenfaces comprise the second domain that is often referred to as the face space.

EIGENFACES eigenfaces

Eigenfaces • Once the eigenfaces have been computed, the system can transform any face image to it's face vector. • Any face image that is to be tested is then transformed to it's face vector representation. Euclidean distance is then used to find the stored face vector that comes closest to the test face vector, and the face is recognized.

Project Evolution THE FACE DETECTION AGORITHM: 1) Collect all gray levels in a long vector u. 2) Collect n samples (views) of each of a person in matrix A (MN X n) 3) Form a correlation matrix L (MN X MN): L=AAT 4) Compute eigen vectors of L, which forms the bases for whole face space. Each face, u, can now be represented as a linear combinationof eigen vectors.

PROJECT EVOLUTION Training: • Create A matrix from training images • Compute C matrix from A. • Compute eigenvectors of C. • Compute eigenvectors of L from eigenvectors of C. • Select few most significant eigenvectors of L for face recognition. • Compute coefficient vectors corresponding to each training image. • For each person, coefficients will form a cluster, compute the mean of cluster.

The image-space and face-space

PROJECT EVOLUTION Recognition: • Create a vector u for the image to be recognized. • Compute coefficient vector for this u. • Decide whether the image is a face or not, based on the distance from the cluster mean.

FUTURE SCOPE Some possible extensions for this project over future years could be: 1)More advanced lighting normalization algorithms before applying eigenfaces method 2)Use of eigenfeatures to increase robustness to change in expression 3)use of multiple subspaces to allow for detection from general perspectives, ie. recognition from profile images 4)A maximum a posteriori (MAP) rule for face detection and recognition which allows the system to be trained not only with positive examples of a face/individual, but also with negative examples