Sonar equipment can produce severe speckle noise interference and degrade image quality when acquiring underwater images due to seafloor reverberation， scattered echoes and equipment. To overcome this challenge， a Non-subsampled contourlet transform （NSCT） sonar image denoising based on neutral ensemble with bilateral filtering is proposed. The method explores a clustering denoising method suitable for dealing with high-frequency subband noise of non-subsampled contourlet transform decomposition， where the neutrosophic set domain is applied as clustering to each layer of high-frequency sub-images， represented by true subsets， uncertainty sets and false sets， to achieve separation of noise and retention of valid information； then bilateral filtering is used to remove the noise mistaken as image information in true subsets； bilateral filtering for low-frequency coefficients smoothing is performed using bilateral filtering. Finally， the processed low-frequency subbands and high-frequency subbands are reconstructed by NSCT to achieve the purpose of removing speckle noise from the image. The experiments show that the proposed method can improve the visual effect of the image， improve the evaluation indexes such as mean square error and peak signal-to-noise ratio， and is also suitable for removing high-density noise.