Face Recognition Using Fusion of Multispectral Imaging

Biometrics has received a lot of attention from the business communities and academic during the last few years. It has emerged as an alternative to traditional forms of identification, like card IDs. Face recognition stands as the one of the most challenging modalities, since it is the natural mode of identification among humans. The Traditional FR methods that based on Visible Spectrum (VS) have facing challenges like pose variation, object illumination, facial disguises, and expression changes. Unfortunately these limitations degrade performance of Face Recognition. To overcome all these limitations, the Infrared Spectrum (IRS) is used in human FR, is chosen due to its inherent nature of being immune towards drastic ambient light changes. This paper presents the design and implement of Human face recognition system based on fusing three spectral images visual face images with near infrared image and thermal images to achieve feature integration in the recognition process and to improve final FR system performance. Gabor wavelet transform is used in feature extraction stage for extracting features. Support Vector Machine (SVM) is used in the classification stage. The proposed approach is tested on a Carl Face Database (Visual/Near Infrared/Thermal) images. Experimental results show that our approach exhibits better detection accuracy for fusion of Multispectral Image (Visual/Near Infrared/Thermal) than visual image. The maximum success of 96.4% recognition has been achieved.