2D-VMD algorithm was used to remove noise to improve image quality.The bearing defect image containing noise was decomposed into finite intrinsic mode function (IMF) components by using 2D-VMD technology;the IMF components were screened by using fuzzy linear exponent and standard deviation to eliminate the noise items to achieve image denoising.The denoising effects of 2D-VMD algorithm,mean and median filtering were compared,and the denoised images were objectively evaluated by using means of mean square error and peak signal-to-noise ratio.The results show that by using 2D-VMD algorithm,the denoising effect is better,and the denoised image can retain more effective information and has better image quality，which can meet the requirements of railway departments for bearing maintenance.