A Novel Sonar Image Denoising Algorithm based on Block Matching

The existing sonar image denoising methods are either based on the spatial domain or the transform domain, which have their own shortcomings. This paper presents an improved BM3D algorithm for sonar image denoising. The parameters are adjusted according to the noise characteristics of sonar image, and a Gaussian filter and gray correction are added before basic estimation in order to remove the speckle noise, and finally its effect is verified on side scan sonar image. The experimental results show that it can effectively eliminate the noise from the side-scan sonar images and retain good structural information.