Abstract: Long coding time is the main problem in fractal image compression in present, to which classification of image blocks is an important and efficient solution. However, In most classification methods, the decreasing of time usually cause quality sacrifice. Therefore we propose a method called “Adaptive Classification based on Two-dimensional Characteristic Vector”. Using first order and second order quadrature of image as main and secondary feature, we obtain a fast method for fractal coding which can get better quality of decoded image