Eye localization for face matching: is it always useful and under what conditions?
CIVR '08 Proceedings of the 2008 international conference on Content-based image and video retrieval
Simultaneous eye tracking and blink detection with interactive particle filters
EURASIP Journal on Advances in Signal Processing
Eye localization in low and standard definition content with application to face matching
Computer Vision and Image Understanding
A novel active heads-up display for driver assistance
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics - Special issue on human computing
Visual-context boosting for eye detection
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
Robust facial landmark detection for intelligent vehicle system
AMFG'05 Proceedings of the Second international conference on Analysis and Modelling of Faces and Gestures
Hi-index | 0.00 |
In this paper we proposed to solve the eye detection and localization problem under a general statistical model based object detection framework. A binary tree representation is used to discover the objects驴 underlying statistical structure. Tree structures enable us to describe the object local statistical structure in a coarse-to-fine fashion. Each subtree explains the statistics for certain local substructure. The tree is built in a top-down fashion. Subsets with negligible conditional independency are found by k-means clustering using mutual information. The conditionally independent features are separated into different subtrees, while more dependent features are tended to appear close in the tree. The distribution of the object can be learned accordingly. Gaussian mixture in the independent component analysis (ICA) subspace is used to model the distribution of each high dependent feature subset, where each independent component explains the local substructure. The use of tree structure enables us to learn the distribution recursively by applying Bayesian criterion. Substantial experiments were done to evaluate the performance over the eyes detection accuracy as well as the localization ability. Experimental results show a better detection accuracy than the Viisage system with a reasonable localization ability, which validate the algorithm.