Aimed at the inaccurate problem of block similarity metric for synthetic aperture radar (SAR) image despeckling using Blockmatching and 3D filtering (BM3D) algorithm, a new SARBM3D despeckling algorithm is proposed based on generalized likelihood ratio criterion (GLR). According to the good results of BM3D algorithm in optical image denoising, based on the transform domain method and nonlocal mind, considering SAR image speckle character, probability similarity measure method is used to block matching thus to achieve better results. The experiment results show that the new algorithm has advantages in the SAR image despeckling performance with higher PSNR and better visual results.