Thermal Image Face Recognition through Transverse Dyadic Wavelet Transform and GLCM

  • N. Jayakumari, Dr. K. Selvam

Abstract

            Face recognition use in biometric application for authentication. Traditional face recognition methods performance is inaccurate in uncontrolled environment, such as illumination and pose. The colour image face recognition method, does not provide accurate recognition. Hence, we propose to use thermal image for face recognition. The face recognition from thermal image recognizes faces with cosmetics and illumination. The feature points of thermally active regions in face image determine by wavelet transform such as Discrete Wavelet Transform (DWT), Non Decimated Stationary Wavelet Transform (NDSWT) and Transverse Dyadic Wavelet Transform (TDyWT). The DWT and NDWT does not show accurate representation of thermal regions in face due to translational invariance and inaccurate wavelet feature coefficient. However, the accurate estimation of approximate and detail coefficient in TDyWT leads to thermal region delineation. The delineated region and feature points spatial pixel relation determine by Grey Level Co-occurrence Matrix (GLCM). The feature points spatial relation between pixels classify by Least Square Support Vector Machine (LSSVM) for face recognition. The robustness of method evaluate with facial images under different illumination, pose conditions and resolution. The TDyWT of face recognition outperforms colour image face method. The TDyWT has 94.73% accuracy for face recognition compared to traditional methods.

Published
2021-10-27
How to Cite
N. Jayakumari, Dr. K. Selvam. (2021). Thermal Image Face Recognition through Transverse Dyadic Wavelet Transform and GLCM. Design Engineering, 8588–8603. Retrieved from http://www.thedesignengineering.com/index.php/DE/article/view/5901
Section
Articles