Self-Supervised SAR Image Denoising Method for Ship Target Detection

In synthetic aperture radar (SAR) images, ship targets are easily submerged by complex background interference, and speckle noise causes blurring of ship edges, severely reducing image quality. This not only affects the usability of the images but also poses challenges for subsequent processing tasks such as image segmentation, target detection, and recognition. To address this issue, this paper proposes a self-supervised SAR image denoising method for ship target detection. This method extracts the regions of interest and suppresses the background in the image using a clustering-based mask. By combining a self-supervised learning strategy, the denoising network is trained only on noisy SAR data without the need for clean reference images, resulting in high-quality images. Experimental verification on GF3 data demonstrates the potential application of the proposed convolutional neural network in SAR image processing.