Face pattern recognition under complex background

We propose a practical method for recognizing the face of acquaintance as a subpattern within a given image, even for the cases where the face is in a crowd, or the face image has a complex background. We assume that the face pattern to be recognized approximately equals an affine transformed (rotated, enlarged, reduced, and translated) input image. In order to detect the face pattern and, at the same time, to estimate the parameters of the affine transformation, we use a Karhunen-Loeve (KL) expansion, spatial correlation, and an approximate equation based on Taylor´s expansion. Moreover, this method also estimates the parameter of lighting to absorb variations of light. Experimental results show that the method is considerably effective