A model for wave-front distortions is built up based on the random phase screen, and the wave-front is fitted by Zernike polynomials with different orders. By comparing and analyzing the power spectrum density (PSD) of the original wave-front and the fitted one, the correspomding relationship between the used orders of Zernike polynomials and the spatial frequency of the wave- front is obtained. Consequently, the insufficient of the conventional method for fitting high frequency phase by Zernike polynomials is revealed, and the local fitting method based on Zernike polynomials is further proposed. In this improvement method, the residual wave-front in the conventional fitting way is decomposed into many subdomains fitted by Zernike polynomials individually. The results indicate that, in the conventional fitting way, the range of spatial frequency of the wave-front that can be accurately described by Zernike polynomials increases gradually with the fitting order of Zernike polynomials, but the increment is modest; even fitted by Zernike polynomials with much higher order, it’s difficult to describe the high frequency components of the wave-front accurately. However, the spatial frequency of wave-front accurately described by Zernike polynomials obviously increases by using the local fitting way. In order to increase the fitting accuracy, increasing the number of subdomains is more superior than increasing the number of the used orders of Zernike polynomials. For the number of subdomains in the local fitting way is fixed, increasing the used orders of Zernike polynomials in subdomains makes more precise fit than increasing the number of overall fitted orders.