Attractor-based person identification

This paper presents a facial identification system which is based on attractors of a set of two dimensional affine transformation of subparts of facial images. In this approach, the attractor models of the known images are calculated and stored in the database. Then, for each known image, its attractor model is iterated from the input image. If the input image and the known image belong to the same person, the iteration of the attractor model produces an image that is similar to the input image. Otherwise, the iteration process changes the content of the input image significantly. Experimental results are reported and discussed.