Abstract:  Image mosaic is a technique of image processing that stitches multiple overlapping images in order to produce a high resolution and global view composition.To widen the view field of sonar image，a novel mosaic method is proposed in this paper.The proposed method mainly includes two steps：sonar image registration and sonar image fusion.It is hard to extract the features from and match the correspondences exactly between sonar images，which have high－level noise and low resolution.Therefore，the FFT method，which is independent on the features of image，is used for sonar image registration.After registration，image fusion is applied to improve the quality of the mosaic result.Image fusion can be categorized into two groups：spatial－based methods and multi－scale transform based methods.The spatial－based fusion techniques are likely to produce blocky artefacts due to the complexities of sonar image.Accordingly，the multi－scale transform based method is more suitable to sonar image fusion.In the fusion step，a multi－scale transform based fusion method is proposed using Non－subsampled Contourlet Transform(NSCT)．First，the source images are decomposed into sub－bands of multi－scale and multi－direction by NSCT.Then，the fusion rules are established to deal with the inherent challenging properties of sonar images，such as featureless，high level noise and low resolution.A threshold－based method is used for merging the coefficients of low－frequency sub－bands while the local contrast is applied to conduct the fusion of high－frequency sub－bands.The fused coefficients of the corresponding sub－bands can be generated by the proposed fusion rules.Finally，the inverse NSCT is implemented on the fused coefficients to generate the fused image.The effectiveness of the proposed method for sonar image mosaic is demonstrated in the experiment.