Online Image Denoising Tool
Reduce random noise using neighborhood averaging. Processing takes place locally in your browser.
Start using
What this tool actually does
Applies a fixed 3 by 3 box-average filter to one browser-decodable image. There are no strength, radius, AI model or noise-profile controls; pressing Generate image averages each RGB channel across the nine neighboring samples.
Checkable behavior
At the image boundary the nearest edge sample is repeated for missing neighbors. Alpha comes from the center source pixel rather than being averaged. A single 255-valued channel surrounded by eight zeros becomes about 28 at the center.
Controls, output and limitations
The method can reduce isolated pixel noise but also softens edges, text and fine texture. It does not distinguish luminance noise from color noise and is not suitable for recovering detail from a severely compressed, blurred or underexposed photo.
The input is capped at 25 MiB and 24,000,000 decoded pixels. Output keeps the decoded width and height and downloads as PNG with -denoise in the filename; EXIF metadata, animation and original encoding are not preserved.
Load one image, press Generate image, compare fine detail at full size and then download only if the smoothing tradeoff is acceptable. Pixel processing is local to the current browser and does not upload the image.
Processing takes place locally in your browser.