Resilient subclass discriminant analysis with application to prelens tear film interferometry

Authors:
Kim L. Boyer;Dijia Wu
Affiliations:
Signal Analysis and Machine Perception Laboratory, Department of Electrical, Computer, and Systems Engineering, Rensselaer Polytechnic Institute, Troy, NY;Signal Analysis and Machine Perception Laboratory, Department of Electrical, Computer, and Systems Engineering, Rensselaer Polytechnic Institute, Troy, NY and Siemens Corporate Research, Princeton ...
Venue:
MCPR'11 Proceedings of the Third Mexican conference on Pattern recognition
Year:
2011

Citing 7
Cited 0

Statistical Pattern Recognition: A Review

IEEE Transactions on Pattern Analysis and Machine Intelligence
Flexible discriminant and mixture models

Statistics and neural networks
Subclass Discriminant Analysis

IEEE Transactions on Pattern Analysis and Machine Intelligence
Discriminant Subspace Analysis: A Fukunaga-Koontz Approach

IEEE Transactions on Pattern Analysis and Machine Intelligence
Rapid and brief communication: Model-based mixture discriminant analysis-an experimental study

Pattern Recognition
Texture based prelens tear film segmentation in interferometry images

Machine Vision and Applications
Markov random field based phase demodulation of interferometric images

Computer Vision and Image Understanding

Quantified Score

Hi-index	0.00

Visualization

Abstract

The study of tear film thickness and breakup has important implications for understanding tear physiology and dynamics. We have developed a complete end-to-end automated system for robust and accurate measurements of the tear film thickness from interferometric video as a function of position and time (following a blink). This paper will primarily address the problem of identifying dry regions on the surface of the contact lens, which is one of the four major components of the system. (The others are motion stabilization, image normalization, and phase demodulation to infer absolute thickness and map the surface. To address the challenging wet/dry segmentation problem, we propose a new Gaussian clustering method for feature extraction in high dimensional spaces. Each class is modeled as a mixture of Gaussians, clustered using Expectation-Maximization in the lower-dimensional Fisher's discriminant space. We show that this approach adapts to a wide range of distributions and is insensitive to training sample size. We present experimental results on the real-world problem of identifying regions of breakup (drying) of the prelens tear film from narrowband interferometry for contact lens wearers in vivo.