NOISE ROBUSTNESS FACE RECOGNITION WITH PHASE INPUT JOINT TRANSFORM CORRELATORS
Pages 149-153
Alin C. TEUSDEA
Abstract
Face recognition implies that the real world scene images are embedded in noise and mix with additive noise therefore it is important to test the correlators in these conditions. Simulations were done on some combinations of amplitude premodulation domains and phase modulation for four input images. Two images have black (null) background and grey (average) background and two images have two kinds of embedding noise. The reference type image from scene image has 50% additive noise for all four test images. These simulations were done to find the combination of the amplitude premodulation domain and phase modulation domain that has the best detection efficiency for all images.