Sonar Image Denoising Based on Anisotropic Guided filtering

The limitations of imaging principle and imaging environment distort the sonar images by various noises, which causes the sonar image to have the characteristics of low resolution and blurred edges. The existing physical models have certain limitations on the description of sonar noise. In order to adapt to a wider range of sonar images, a filter with blind denoising is the best choice; an edge-preserving filter effectively solves the problem of blurred edges; in real application scenarios, processing efficiency is one of the important factors, and a filter with a linear model has a high processing efficiency. In this paper, the anisotropic guided filter is applied to sonar image denoising. Firstly, the adaptive regularization parameters are obtained using the local variance and the approximate median variance of the natural image. Secondly, the local mean, local variance, and regularization parameters are used to obtain the parameters of the guided filter model. Thirdly, anisotropic weighting is applied to the guided filter model parameters. Finally, the output image is obtained by integrating the anisotropic weighted model parameters with the guidance image. The experiments show that the anisotropic guided filter outperforms the comparison algorithm in the removal of both synthetic and sonar image noise, and achieves efficient sonar image denoising and edge preservation.