Non-local Means Filtering (NLMF) has been a popular issue in the image filtering field.The existing NLMFs based pre-selections are analyzed

and it is pointed out that they all have deficiencies in terms of feature extraction from image patches.An adaptive and effeicient NLMF method is proposed using singular value decomposition (SVD) in the gradient domain.Our contributions to NLMF based pre-selection are:1)the robust pre-selection method based structure feature from image patch;2)the relation between size of the similar sets and filtering performance is analyzed;3)automatic selection of similar patches;4)local adaptive selection of the filtering parameter.In addition

the symmetry of the Euclidean distance is considered to accelerate the proposed method further.The experimental results show that the proposed method outperforms the original NLMF and other fast NLMFs on subjective and objective aspects

and has rapid running speed.The proposed method is an efficient filtering method.