IMAGE PREPROCESSING FOR CLASSIFICATION BIOMETRIC IDENTIFICATION BY A
IMAGE PRE-PROCESSING FOR CLASSIFICATION (BIOMETRIC IDENTIFICATION) BY A NEURAL NETWORK Anthony Vannelli, Steve Wagner, and Ken Mc. Garvey Horizon Imaging, LLC email: info@horizonimaging. com Horizon Imaging, LLC Innovative Solutions in Image Processing
Neural Network Preprocessor and Classifier Raw 512 x 480 Image Reduced data set Neural Preprocessor Neural Network Classifier Classification Output • Wavelets • Feed-forward Network • PCA • Back-propagation Training • Image “Zones” • Single Hidden Layer • Combining Networks Horizon Imaging, LLC Innovative Solutions in Image Processing
Curse of dimensionality • 512 x 480 raw image or 245, 760 inputs to network • Large neural network • Poor classification performance • Slow convergence Horizon Imaging, LLC Innovative Solutions in Image Processing
Biometric Identification Region of Interest 320 x 160 = 51, 200 pixels Horizon Imaging, LLC Innovative Solutions in Image Processing
Preprocessing Techniques • Non-parametric • “Holistic” • Data-driven • No Hand Geometry • No Fidiucial Points Horizon Imaging, LLC Innovative Solutions in Image Processing
Preprocessing Techniques • • Principal components • Large eigen-values help to classify • Reduces dimensionality Image Processing Zones • Divide and conquer • 2 x 2 zones (160 x 80 pixels) • 4 x 4 zones (80 x 40 pixels) • Ensemble of neural networks Horizon Imaging, LLC Innovative Solutions in Image Processing
Preprocessing Techniques • Combining Neural Networks • Pick the network with the “best fit” • Average the network outputs • Voting Scheme Horizon Imaging, LLC Innovative Solutions in Image Processing
Voting Scheme to Combine Networks Neural Net #1 y 1 > T Neural Net #2 Input Vector y 2 > T 1 2 Combined Output = N Neural Net #N y. N > T 0 for yi T i = 1 for yi > T Figure 3. Voting scheme to combine Neural Networks Horizon Imaging, LLC Innovative Solutions in Image Processing i
Preprocessing Technique using Wavelets • Coiflet wavelet • Daubechies wavelet • Haar wavelet (averages adjacent pixels) Second-level wavelet approximation Horizon Imaging, LLC Innovative Solutions in Image Processing
One-Level of a Wavelet Transform Low 2 LL High 2 LH Low 2 HL High 2 HH 2 Image f(x, y) High 2 Horizontal filter Horizon Imaging, LLC Innovative Solutions in Image Processing Vertical filter
Third-level Wavelet Decomposition LL HL LH HH Horizon Imaging, LLC Innovative Solutions in Image Processing
Test Case with Single Classifier 320 x 160 pixels 512 x 480 Image Preparation Wavelet Transform PCA Figure 7. Test case with single classifier Horizon Imaging, LLC Innovative Solutions in Image Processing Neural Classifier Output
Test Case with Multiple Classifiers Image 1 512 x 480 Image Preparation Neural Classifier Wavelet Transform Combine Networks 320 x 160 pixels Image N Neural Classifier Figure 8. Test case with multiple classifiers Horizon Imaging, LLC Innovative Solutions in Image Processing Output
Test Cases A. Coiflet 6 -coefficient wavelet to 3 levels; 3 rd level approximation image (40 x 20 pixels) and 3 sidebands form input to 4 neural networks with 800 inputs each. B. Daubechies 6 -coefficient wavelet to 3 levels; 3 rd level approximation image (40 x 20) and 3 sidebands form input to 4 neural networks with 800 inputs each. C. Coiflet 6 -coefficient wavelet to 2 levels (80 x 40 pixels); 4 image zones fed to 4 separate neural networks with 800 inputs each. Horizon Imaging, LLC Innovative Solutions in Image Processing
Test Cases D. Daubechies 6 -coefficient wavelet to 2 levels (80 x 40 pixels); 4 image zones fed to 4 separate neural networks with 800 inputs each. E. Harr wavelet to 2 levels (80 x 40 pixels); 4 image zones fed to 4 separate neural networks with 800 inputs each. F. Harr wavelet to 2 levels (80 x 40 pixels) and then PCA transform fed to a neural network with 512 inputs. Horizon Imaging, LLC Innovative Solutions in Image Processing
Test Cases G. Harr wavelet to 3 levels (40 x 20 pixels) fed to a neural network with 800 inputs. H. Coiflet 6 -coefficient wavelet to 1 level (160 X 80 = 12800 pixels). The first level approximation image is divided into 16 image zones (40 x 20 pixels per zone). The zones are fed into separate neural networks with 800 inputs each. Horizon Imaging, LLC Innovative Solutions in Image Processing
Summary of Performance Horizon Imaging, LLC Innovative Solutions in Image Processing
- Slides: 17