A No-Reference Image Quality Assessment For Detecting Illumination Alteration
Cerine Tafran, Mohamad El-Abed, Islam Elkabani, Ziad Ahmad Osman · 2019
The Quality assessment of a face image is a topic of great interest for biometric applications where images with bad samples decrease the system performance and increase authentication errors, especially in biometric passport applications that use only a single image for enrollment. Thus, in order to have a useful biometric authentication system, the quality of the biometric sample images must be controlled. This paper presents a no-reference quality assessment method which detects the illumination problem using Symmetric Based Features along with BLIINDS Based Features. The experimental results on the AR database recorded an accuracy of 90.3 % by using Stochastic Gradient Descent (SGD) classifier.