The AFM image will be distorted due to the tip-broadening effect. It may be mathematically represented as the effect of the tip’s convolution on the real image. As a result, the true geometry of the sample surface can not be obtained. If the tip shape could be known accurately, such distortion can be eliminated with the adoption of disconvolution operation. It is of practical significance in the accurate reconstruction of the AFM image. Among the current tip modeling algorithms, the blind tip estimation algorithm based on mathematical morphology is widely applied. However, it takes much time and has difficulties in defining the optimal noise threshold. In view of the above problems, this paper proposes a new blind tip evaluation method to accelerate the speed of blind tip modeling calculation and realize the accurate reconstruction of the AFM image. The simulation and the experimental results demonstrate the feasibility and effectiveness of the above research method.