Adaptive Guided Filtering Based BM3D Sonar Image Denoising Algorithm

Forward-Iooking sonar (FLS) is one of the essential imaging equipment used in exploring underwater targets. However, image quality has a significant impact on target detection. FLS is typically disturbed by internal circuit noise and external noise, resulting in images with low resolution, low contrast, fuzzy target edges, and other features. This paper proposes an adaptive guided filtering based BM3D sonar image denoising algorithm (AGF-BM3D) to deal with the denoising problem of FLS. The paper uses a FLS working at a center frequency of 750 kHz in a shallow area for data acquisition. The forward-looking sonar and the target are positioned at a uniform depth of 3m below the water surface, with a separation of 42m, to achieve optimal acoustic imaging performance. A total of 116 sonar images were acquired during the experiment for subsequent FLS image denoising. Applying this method with the classical denoising algorithm to the shallow sea real data, the results show that AGF-BM3D is able to remove the electrical and environmental noise from the FLS image while keeping the edges well.