Based on the limitation analysis of conventional de-convolution approaches, an improved beam sharpening approach based on constant false alarm rate (CFAR) and de-convolution is proposed to deal with the poor resolution problem in azimuth direction of airport surface movement radar. Using the advantage of CFAR, the proposed method can adjust the processing threshold of de-convolution adaptively to avoid the uselessly iterative solving process in the noise region. Due to this improvement, it ensures the beam sharpening of targets in the interesting region and suppresses the occurrence of false targets simultaneously. At last the false rate of the airport surface movement radar (ASMR) system can be reduced effectively. The computer simulation analysis and real data processing show that our proposed algorithm can successfully solve the noise sensitivity in the conventional de-convolution approaches, and has important engineering application value.