BYTETOOLS

Reduce Image Noise

Clean grain and speckle out of a photo with a median or edge-preserving bilateral filter, applied to all channels or to colour noise only.

Drop a grainy or speckled photoProcessed locally in your browser — never uploaded

What is the Reduce Image Noise?

High-ISO shots, old scans and phone photos taken in dim light all pick up noise: random speckle in the detail and blotchy colour patches in flat areas.

  • 3 × 3 and 5 × 5 median filters for salt-and-pepper and sensor speckle
  • Bilateral filter with independent spatial and range sigma, built on precomputed weight tables
  • Chroma-only mode that removes colour blotches without touching detail
  • Strength slider that blends the filtered result back against the original
  • Explicit working-resolution cap so large photos cannot freeze the tab
  • Runs entirely in your browser — the photo is never uploaded

How to use the Reduce Image Noise

  1. 1

    Drop in the noisy photo.

  2. 2

    Choose a filter: 3 × 3 median for light speckle, 5 × 5 median for heavier grain, or bilateral for edge-preserving smoothing.

  3. 3

    Set what it applies to — all RGB channels, chroma only to kill colour blotches, or luma only.

  4. 4

    For the bilateral filter, raise the spatial sigma to smooth wider and the range sigma to let it cross bigger tone differences.

  5. 5

    Pull the strength slider back if the result looks plastic, then click Download PNG.

About the Reduce Image Noise

High-ISO shots, old scans and phone photos taken in dim light all pick up noise: random speckle in the detail and blotchy colour patches in flat areas. This tool offers the two filters that actually work on it — a median filter, which throws away isolated outlier pixels, and a bilateral filter, which averages neighbours only when they are similar in tone so edges survive the smoothing.

The most useful control is what the filter is applied to. Colour noise lives almost entirely in the chroma channels, so filtering chroma only removes the blotches while leaving every scrap of luminance detail intact. Filtering luma only softens grain but keeps the colour crisp, and filtering all three RGB channels does both at once.

Neighbourhood filters cost one pass per tap, so the photo is resampled to your chosen working resolution before filtering — a 12-megapixel original through a 5 × 5 median would be hundreds of millions of comparisons. Nothing is uploaded; all of it runs on a canvas in your browser.

Frequently asked questions

What is the difference between a median and a bilateral filter?

A median filter replaces each pixel with the middle value of its neighbours, which deletes isolated specks outright. A bilateral filter takes a weighted average but ignores neighbours whose tone is far from the centre pixel, so it smooths flat areas while leaving edges sharp.

Why does chroma-only denoising keep more detail?

Your eye reads detail mostly from brightness and colour mostly from large areas, and sensor colour noise shows up as blotches in the chroma channels. Smoothing chroma alone erases those blotches while the luminance channel, which carries the sharpness, is left completely untouched.

Does noise reduction make images blurry?

Any denoiser trades some detail for smoothness. The bilateral filter is designed to minimise that by refusing to average across edges, and the strength slider lets you dial the effect back to a partial blend if the result starts looking waxy.

Why is the downloaded image smaller than my original?

Neighbourhood filters get expensive fast, so the photo is resampled down to the working resolution you choose before filtering, and the download is that filtered image. Pick a higher working resolution if you need more pixels and can wait longer.

Which settings work best for scanned film grain?

Start with a 3 × 3 median on all channels at about 60% strength. If the grain is coloured rather than grey, switch to chroma only — that usually removes the distraction while keeping the film look you actually want.

Related tools