The resolution of space object images observed by ground-based telescope is greatly limited due to the influence of atmospheric turbulence. An improved blind deconvolution method is presented to enhance the performance of turbulence degraded images restoration. Firstly, a mixed noise model based blind deconvolution cost function was deduced under Gaussian and Poisson noise contamination of measurement. Then, point spread function (PSF) was described by wavefront phase aberrations in the pupil plane according to Fourier Optics theory. In this way, the estimation of PSF was generated from the wavefront phase parameterization instead of pixel domain value. Finally, the cost function was converted from constrained optimization problem to non-constrained optimization problem by means of parameterization of object image and PSF. Experimental results show that the proposed method can recover high quality image from turbulence degraded images effectively. 