Dynamic Convolutional Network with Noise Gating for Enhanced Image Denoising
K P Aswathi, Girish Jadhav, D. Nirmal Raj, Ashok Chanabasangouda Patil · 2025
Image denoising plays a vital role in enhancing image quality across various applications, including medical imaging, autonomous navigation, and multimedia. Traditional denoising methods often struggle to handle complex noise patterns and fail to preserve intricate image details, especially under varying noise conditions. To address these challenges, this paper introduces a novel Dynamic Convolutional Network with Noise Gating (DCNNG) designed to adaptively suppress noise while maintaining critical image structures. The proposed DCNNG employs a dynamic convolutional layer that adjusts its kernel weights based on the statistical characteristics of the input noise. Additionally, a noise gating mechanism is integrated to selectively emphasize regions with higher noise, ensuring focused denoising and minimal loss of image details. The model leverages an optimized loss function combining mean squared error (MSE) and structural similarity index measure (SSIM) to balance denoising performance and detail preservation. Experimental evaluations on standard datasets demonstrate that DCNNG achieves superior performance compared to state-of-the-art methods, consistently delivering higher peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) values across varying noise levels. The model’s robustness and adaptability are further validated under different noise distributions, showcasing its practical applicability in realworld scenarios. These results highlight DCNNG’s potential as a reliable solution for advanced image denoising tasks, paving the way for future research in dynamic neural networks for image processing.