Aplica

1
University of São Paulo Institute of Mathematics
and Statistics Department of Computer
Science Tracking Facial Features Using Gabor
Wavelet Networks
Rogério Schmidt Feris Roberto Marcondes Cesar
Junior rferis,cesar_at_ime.usp.br
2
Outline

• Motivation
• Wavelet Transforms
• Face Representation Using GWNs
• Facial Feature Tracking
• Experimental Results
• Future Work

3
Motivation

• Face Recognition from Video Sequences
• Problem Assumptions

4
Wavelet Transforms

• Continuous Wavelet Transform
• Discrete Wavelet Transform
• Wavelet Networks

5
Face Representation Using GWNs

• Mother Wavelet Odd-Gabor Function

6
Face Representation Using GWNs
7
Face Representation Using GWNs
Original Image
Wavelet Representation
Wavelet parameters are chosen from the continuous
parameters space !!
8
Face Representation Using GWNs

• Progressive Attention

Interest
32, 52, 100 and 320 wavelets

• Direct Calculation of Weights

9
Facial Feature Tracking

• Repositioning

10
Facial Feature Tracking

• Superwavelet

• Reparametrization

11
Facial Feature Tracking

• Initialization
• Repositioning in each Frame
• Facial Feature Tracking

12
Experimental Results

• Efficiency and Robustness

http//www.ime.usp.br/rferis
13
Future Work

• Tracking Based on Wavelet Weights
• Face Detection Using GWNs
• 3D Wavelet Model

14
Questions ?
Rogério Schmidt Feris
rferis_at_ime.usp.br
http//www.ime.usp.br/rferis
15
Experimental Results

• Efficiency and Robustness

http//www.ime.usp.br/rferis