In order to restore turbulence-degraded images exactly and rapidly, an iterative blind deconvolution (IBD) algorithm in the frequency domain based on temporal signature is proposed. The temporal signature regularization and Tichonov regularization are incorporated in the cost function. The constraints of non-negativity, energy and bandwidth of the PSFs are added in the iterative blind deconvolution to estimate the object image and point spread functions (PSFs) by the second order conjugation gradient (CG) optimization method. Structure-adaptive applicability filter is used to reduce noise and promote the edges of images. The experimental results show that the proposed algorithm is efficient to recover different intensity turbulence-degraded images and robust with high noise-resisting ability.