Most of the existing deblurring algorithms rely on regularization technique and achieve the best estimation of the latent image by approximating t he optimal objective function cost iteratively.As lack of iteration termination cr iterion (ITC), these algorithms are usually implemented with fixed iteration steps,which leads to low efficiency and poor quality of the deblurred image in many case s. Considering the fact that the residual image between the intermediate estimated image convoluted with blur kernel and the blurred image at each step has signi ficant characteristics of Gaussian distribution,in this paper,we employ the general ized Gaussian distribution (GGD) model to characterize residual image for each iterative step,and the GGD model parameter is selected as a deblurring measure (DM).Based on the DM,an ITC is designed to terminate the iterative process adaptively to improve the computational efficiency of deblurring algorithm under the premise of guaranteeing the image quality.Extensive experiments conducted by the classical nonlocally centralized sparse representation (NCSR) algorithm show that for motion blur,Gaussian blur,and out of focus blur,the proposed DM metric and th e ITC are effective to determine whether the estimated image obtained at each iterative step achieves the optimal quality,greatly improving the computational efficiency of NCSR algorithm.Moreover,the ITC proposed i n this paper can be applied to other iterative deblurring algorithms after adjusting the relevant parameters.