Hierarchical Fusion of Multi Spectral Face Images for Improved ... - IIIT
Hierarchical Fusion of Multi Spectral Face Images for Improved ... - IIIT
Hierarchical Fusion of Multi Spectral Face Images for Improved ... - IIIT
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Short Wave IR<strong>Face</strong> ImageAmplitudeFeaturesImage<strong>Fusion</strong>usingDWTFused <strong>Face</strong>ImageSVMLearningbasedFeature<strong>Fusion</strong>Fused FeatureVector <strong>for</strong><strong>Face</strong> RecognitionLong Wave IR<strong>Face</strong> ImagePhaseFeaturesLevel1-Image<strong>Fusion</strong>Level2-Feature <strong>Fusion</strong>Fig. 6. Two-level hierarchical fusion <strong>of</strong> multi spectral face images4 Experimental ResultsExperiments are per<strong>for</strong>med on the face database from Equinox Corporation [29]. Thedatabase contains images captured in long-wave, medium-wave and short-wave infraredlight along with the images in the visible spectrum. Visible spectrum captures theelectromagnetic energy in the range 0.4 - 0.7 µ m, and infrared spectrum comprises <strong>of</strong>thermal and reflectance wavebands in the range <strong>of</strong> 0.7 - 14.0 µ m. Short-wave infrared(SWIR) which is in the range <strong>of</strong> 0.9 - 2.4 µ m is reflective. Thermal IR spectrum isdivided into two parts, medium-wave infrared (MWIR) from the spectral range <strong>of</strong> 3.0 -5.0 µ m and long-wave infrared (LWIR) from 8.0 - 14.0 µ m. Emitted radiation dependson the temperature and emissivity <strong>of</strong> the material. There are images from 91 individualsin the database with variations in facial expression, illumination, and occlusion due toglasses <strong>for</strong> all four light spectrums. The results are divided into two parts. First part20