Abstract: Noise image must be smoothed before segmentation,and the filtering result relies on the choice of parameters when bilateral filtering is used to smooth images.In order to reduce difficulties in parameter selections,an MSBF(multi-scale bilateral filtering) algorithm is proposed to synthesize the results of bilateral filtering under different scales.Firstly,MSBF is used to the smoothing pretreatment of the noise image,and the edge can be well retained at the same time of denoising. Then discontinuity-adaptive Gaussian Markov random field(DA-GMRF) model is employed to segment the image.Results of segmentation experiments on several images show that the proposed algorithm is effective.