In this paper，a fractal based graph theoretic approach for face recognition is presented.Based on the image compression of fractal

pixel blocks(Range blocks) are defined as basal elements. The interdependence of pixel blocks is inherent within the fractal code in the form of circular plants. Fristly，the fractal code of face images is calculated and the corresponding circular plants can be obtained.Secondly，affine parameters for each pixel block are computed according to the acquired circular plants. Lastly，face recognition can be realized by appropriate distance measurement. The presented way is more effective，faster and more robust than the previous technique based on pixels. In addition，we detail the growth of circular plants in a general case to cover the shortage of predecessor.