GPR Image Denoising Based on Improved DNCNN

Ground penetrating radar (GPR) images are often affected by Gaussian noise, which reduces their clarity. While traditional filtering methods can mitigate noise to some extent, they often struggle to preserve image details simultaneously. Deep learning-based Denoising Convolutional Neural Network (DNCNN) has demonstrated notable performance in image denoising, yet its effectiveness in handling Gaussian noise in GPR images leaves room for improvement. This study enhances the DNCNN model by integrating a Bottleneck Attention Module (BAM) module, enabling the network to better identify and focus on both noise and image details, thereby improving denoising performance. Experimental results show that the modified DNCNN model not only surpasses traditional techniques but also outperforms the original DNCNN in GPR image denoising. The improvements in PSNR and SSIM metrics confirm the effectiveness of the proposed approach<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[3]</sup>.