Deep learning-based non-local means CCD image denoising
Jialong Cai, Tao Jiang, Shilin Duan · 2025
A deep learning-based Non-Local Means (NLM) approach for the denoising of CCD images is proposed in this paper. A four-layer neural network architecture is proposed, in which the parameters are meticulously optimized through backpropagation, utilizing pixel neighborhood similarity as the input feature. The integration of deep learning with Non-Local Means (NLM) is validated through experiments that reveal superior denoising capabilities and enhanced detail preservation when contrasted with conventional NLM and median filtering techniques. For broader applicability, further enhancements in smoothness and data augmentation are essential.