# Self-Supervised training for blind multi-frame video denoising

arXiv: https://arxiv.org/abs/2004.06957 · HTML: https://mlsanity.com/paper/2004.06957/

## Key facts

- Cited by 26 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 15 Apr 2020.

## Built on by (most cited first, 26 of 26)

- [Imposing Consistency for Optical Flow Estimation](https://mlsanity.com/paper/2204.07262/) (2022)
- [Self-Supervised Super-Resolution for Multi-Exposure Push-Frame Satellites](https://mlsanity.com/paper/2205.02031/) (2022)
- [RViDeformer: Efficient Raw Video Denoising Transformer with a Larger Benchmark Dataset](https://mlsanity.com/paper/2305.00767/) (2023)
- [Recurrent Self-Supervised Video Denoising with Denser Receptive Field](https://mlsanity.com/paper/2308.03608/) (2023)
- [BVI-RLV: A Fully Registered Dataset for Low-Light Video Enhancement](https://mlsanity.com/paper/2407.03535/) (2024)
- [Temporal As a Plugin: Unsupervised Video Denoising with Pre-Trained Image Denoisers](https://mlsanity.com/paper/2409.11256/) (2024)
- [Learning Task-Oriented Flows to Mutually Guide Feature Alignment in Synthesized and Real Video Denoising](https://mlsanity.com/paper/2208.11803/) (2022)
- [Video Dynamics Prior: An Internal Learning Approach for Robust Video Enhancements](https://mlsanity.com/paper/2312.07835/) (2023)
- [BVI-Lowlight: Fully Registered Benchmark Dataset for Low-Light Video Enhancement](https://mlsanity.com/paper/2402.01970/) (2024)
- [Spatiotemporal Blind-Spot Network with Calibrated Flow Alignment for Self-Supervised Video Denoising](https://mlsanity.com/paper/2412.11820/) (2024)
- [Self-supervision versus synthetic datasets: which is the lesser evil in the context of video denoising?](https://mlsanity.com/paper/2204.11493/) (2022)
- [Coarse-to-Fine Video Denoising with Dual-Stage Spatial-Channel Transformer](https://mlsanity.com/paper/2205.00214/) (2022)
- [L1BSR: Exploiting Detector Overlap for Self-Supervised Single-Image Super-Resolution of Sentinel-2 L1B Imagery](https://mlsanity.com/paper/2304.06871/) (2023)
- [Unsupervised Coordinate-Based Video Denoising](https://mlsanity.com/paper/2307.00179/) (2023)
- [Deep learning based domain adaptation for mitochondria segmentation on EM volumes](https://mlsanity.com/paper/2202.10773/) (2022)
- [Unsupervised Microscopy Video Denoising](https://mlsanity.com/paper/2404.12163/) (2024)
- [Classic Video Denoising in a Machine Learning World: Robust, Fast, and Controllable](https://mlsanity.com/paper/2504.03136/) (2025)
- [Stabilize, Decompose, and Denoise: Self-Supervised Fluoroscopy Denoising](https://mlsanity.com/paper/2208.14022/) (2022)
- [Unsupervised Methods for Video Quality Improvement: A Survey of Restoration and Enhancement Techniques](https://mlsanity.com/paper/2507.08375/) (2025)
- [Joint denoising and HDR for RAW video sequences](https://mlsanity.com/paper/2201.07066/) (2022)
- [Towards Real-World Video Denosing: A Practical Video Denosing Dataset and Network](https://mlsanity.com/paper/2207.01356/) (2022)
- [SelfReDepth: Self-Supervised Real-Time Depth Restoration for Consumer-Grade Sensors](https://mlsanity.com/paper/2406.03388/) (2024)
- [Combining Pre- and Post-Demosaicking Noise Removal for RAW Video](https://mlsanity.com/paper/2410.02572/) (2024)
- [Frames2Residual: Spatiotemporal Decoupling for Self-Supervised Video Denoising](https://mlsanity.com/paper/2603.10417/) (2026)
- [Self-Supervised Uncertainty Estimation For Super-Resolution of Satellite Images](https://mlsanity.com/paper/2603.14074/) (2026)
- [Enhanced Self-Supervised Multi-Image Super-Resolution for Camera Array Images](https://mlsanity.com/paper/2604.06816/) (2026)

## Abstract (arXiv)

We propose a self-supervised approach for training multi-frame video denoising networks. These networks predict frame t from a window of frames around t. Our self-supervised approach benefits from the video temporal consistency by penalizing a loss between the predicted frame t and a neighboring target frame, which are aligned using an optical flow. We use the proposed strategy for online internal learning, where a pre-trained network is fine-tuned to denoise a new unknown noise type from a single video. After a few frames, the proposed fine-tuning reaches and sometimes surpasses the performance of a state-of-the-art network trained with supervision. In addition, for a wide range of noise types, it can be applied blindly without knowing the noise distribution. We demonstrate this by showing results on blind denoising of different synthetic and realistic noises.