Learned nonlocal feature matching and filtering for RAW image denoising

Most modern image denoising networks rely on large depth and parameter counts, making limited use of technical knowledge brought by classical model-based denoisers. This poses a challenge not only for understanding their properties, but also for deploying them on real devices which may present resource constraints and diverse noise profiles. Tackling both issues, we propose a RAW-to-RAW denoising architecture that incorporates the structure of classical self-similarity-based denoisers into a fully learnable neural network. Our design centers on a novel nonlocal block that parallels the established pipeline of neighbor matching, collaborative filtering and aggregation popularized by nonlocal patch-based methods, operating on learned multiscale feature representations. This built-in nonlocality efficiently expands the receptive field, so that a single block per scale with a moderate number of neighbors is sufficient to obtain high-quality results. Training the network on a curated dataset with clean real RAW data and modeled synthetic noise while conditioning it on a noise level map yields a sensor-agnostic denoiser that generalizes effectively to unseen devices. Both quantitative and visual results on benchmarks and in-the-wild photographs position our method as a practical solution for real-world RAW denoising, achieving results competitive with state-of-the-art convolutional and transformer-based denoisers while having significantly fewer parameters and a faster runtime. The code is available at https://github.com/MIA-UIB/nonlocal-matchfilter.