ffmpeg-denoise-restore

FFmpeg denoise filters like hqdn3d, nlmeans, bm3d, atadenoise, and fftdnoiz clean noise from archival footage, webcams, and low-light captures. A provided Python driver script executes these filters to process input files and output restored video.

FFmpeg upscales low-resolution video using dnn_processing with ESPCN, EDSR, or SRCNN models. This AI-based super-resolution recovers lost detail in low-resolution clips during the video restoration pipeline.

Applying FFmpeg spatial and temporal denoise filters like nlmeans or bm3d removes film grain and noise from archival footage. These filters target grain while preserving detail, often combined with audio cleanup for full restoration.

FFmpeg supports audio denoise for noisy recordings using afftdn, anlmdn, and arnndn filters. These can be combined with video denoise filters to create end-to-end pipelines that clean both audio and video simultaneously.

For webcam captures, hqdn3d and atadenoise are effective FFmpeg filters to clean noise. A medium-denoise pass can be run via the included Python driver script by specifying an input file and output path.