A Brief Survey onFace Recognition Systems Amir Omidvarnia March 2007
Outline • Introduction • Face Recognition Concepts • 2D Face recognition Systems • An Example • 3D Face Recognition • Suggestions
Introduction • What is face recognition? • Applications • Security applications • Image search engine
Requirements • Accurate • Efficient • Light invariant • Rotation invariant
Applications of Face Recognition • Desire to locate specific individuals • Criminals • TERRORISTS • Missing Children • Surveillance
Face Recognition Concepts • Enrollment An initial featureset is constructed from the relevant physical traits of the user.
Face Recognition Concepts • Verification Extracted featureset from each person is compared with the enrollment feature set. If the resulting score value is above a predefined threshold, the user is considered to be authenticated.
Face Recognition Concepts • Identification In contrast to the verification use case, with identification the (claimed) identity of the user is not known in advance, but shall be determined based on sample images of the user's face and a set (population) of feature sets with known identities.
An Example:FaceVACS Architecture • Face Localization • Eye Localization • Image Quality Check • Normalization • Preprocessing • Feature Extraction • Construction of the Reference Set • Comparison
Face Recognition Concepts • The Facial Identification Record (FIR) In the result of processing the raw samples (images), e.g. during enrollment, feature sets are created. In the context of FaceVACS-SDK we use the term FIR for these feature sets.
Face Recognition Concepts • Enrollment and Verification
Face Recognition Concepts • FAR, FRR and EER FAR (False Acceptance Rate) is the probability that a sample falsely matches the presented FIR. FRR (False Rejection Rate) is the probability that a sample of the right person is falsely rejected. The value of FAR and FRR at the point where the plots cross is called the Equal Error Rate (EER).
Face Recognition Concepts • FAR, FRR and EER
Face Recognition Systems • Feature-Based • Appearance-Based • Model-Based
Feature-Based Algorithms • Geometric Features • Texture • Skin color • Multiple features
Appearance-Based Algorithms • Eigenface • Fisherface • SVM • Neural Networks • Hidden Markov Models
Model-Based Algorithms • Face Bunch Graph • Predefined face templates • Deformable templates
An Example • Sample Image
An Example • Eye location found by the algorithm
An Example • After Normalization
An Example • After Preprocessing
An Example • Extracting local features
An Example • Forming the reference set of the image
Image Database • Effective Factors in combining FIRs • Influence and arrangement of lighting conditions • Sample Quality • Orientation of Samples • Adornment • Face Angle • Face Appearance
Suggestions • A comprehensive Survey on 2D Face Recognition Algorithms • Face Detection • Face Segmentation • Feature Extraction • Facial Models • Texture Analysis • Towards 3D Face Recognition • Combining other Biometrics such as Iris Recognition Towards Multimodal Systems