Using Face Quality Ratings to Improve Real-Time Face Recognition

Authors:
Karl Axnick;Ray Jarvis;Kim C. Ng
Affiliations:
Monash University, Clayton, Australia 3800;Monash University, Clayton, Australia 3800;Monash University, Clayton, Australia 3800
Venue:
PSIVT '09 Proceedings of the 3rd Pacific Rim Symposium on Advances in Image and Video Technology
Year:
2009

Citing 9
Cited 0

Detecting Faces in Images: A Survey

IEEE Transactions on Pattern Analysis and Machine Intelligence
Face Recognition Using Line Edge Map

IEEE Transactions on Pattern Analysis and Machine Intelligence
Real-Time Face Recognition Using Feature Combination

FG '98 Proceedings of the 3rd. International Conference on Face & Gesture Recognition
Mugshot Database Acquisition in Video Surveillance Networks Using Incremental Auto-Clustering Quality Measures

AVSS '03 Proceedings of the IEEE Conference on Advanced Video and Signal Based Surveillance
The CMU Pose, Illumination, and Expression Database

IEEE Transactions on Pattern Analysis and Machine Intelligence
Video-Based Framework for Face Recognition in Video

CRV '05 Proceedings of the 2nd Canadian conference on Computer and Robot Vision
Face Identification by a Cascade of Rejection Classifiers

CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops - Volume 03
Complete Two-Dimensional PCA for Face Recognition

ICPR '06 Proceedings of the 18th International Conference on Pattern Recognition - Volume 03
Face recognition using artificial neural network group-based adaptive tolerance (GAT) trees

IEEE Transactions on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

A Face Quality Rating (FQR) is a value derived from a face image that indicates the probability that the face image will be successfully recognized by a specific face recognition method. The FQR can be used as a pre-filter in real-time environments where thousands of face images can be captured every second by multiple surveillance cameras. With so many captured face images, face recognition methods need to strategically decide which face images to attempt recognition on, as it is prohibitively difficult to attempt recognition on all of the images. The FQR pre-filter optimizes processor time utilization resulting in more people being recognized (faster and more accurately) before they leave the surveillance cameras' views. We generate FQR values using Multiple Layered Perceptron (MLP) neural networks. We then use these MLPs in a real-time environment to experimentally prove that FQR pre-filtering improves the speed and accuracy of any real-time face recognition method...