Considering that during the scanning image of an atom force microscopy（ AFM）, the AFM tip shape convolves with the surface morphology, and the feature in the surface is broaden in the image ,thus the tip broaden effect will lead to distortion of the image, a study on reconstructing the tip shape to decrease the image distortion by using de- convolution operation was conducted. The blind tip estimation, a method which is often used to evaluate the tip shape currently, was analyzed. The analysis indicated that though this blind tip estimation method does not require the calibration film whose surface morphology is pre-known, such as porous aluminum, but it makes the noise thresh- old influence on the tip estimation result greatly, and the surface of the calibration film is easily contaminated. These disadvantages could reduce the efficiency of the method. As for these problems, the mathematic morphology based method was proposed to estimate the tip shape through imaging the regular spherical nano-particle on the fiat CD surface, then achieve the image reconstruction. The simulation and experimental results show the effectiveness of the proposed method.