Clean up grain and speckles from high-ISO or scanned photos with four denoising methods, one strength slider, and a before-and-after compare.
Replaces each pixel with an average of its neighbours, chosen so that random noise cancels out while real structure stays. The four methods differ in how they pick the neighbours: a plain blur uses all of them, the median takes the middle value, the bilateral filter weights pixels by how similar their colour is, and non-local means searches for whole patches that look alike. Strength runs from 1 to 10 and controls the radius or weighting for whichever method is active.
A 4000 × 3000 phone photo taken indoors at ISO 3200 has coloured grain across the walls and ceiling. Smart smooth at strength 5 clears the walls in about two seconds and keeps the edge of the door frame sharp; the JPEG afterwards is about 40 percent smaller because noise is expensive to compress. A scanned 1980s print at 1800 × 1200 with white specks of dust responds better to Speckle at strength 3, which removes single dots without touching the grain that gives the print its character.
Use it on high-ISO night shots, old scans, heavily compressed downloads and screenshots of video, ideally before Sharpen Image and before Image Compressor so the encoder is not wasting bytes on grain. Use a light strength on portraits or skin starts to look plastic. Skip it on images that are already clean, on pixel art, and on text-heavy screenshots where Speckle would eat thin strokes. If you actually want grain, the opposite tool is Add Noise.
Smart smooth is a bilateral filter: strength sets both the spatial radius and how different a neighbour may be before it stops counting. Speckle is a median filter with a radius of about a third of the strength, ideal for dust and dead pixels. Soft blur is a Gaussian blur of that many pixels, fast but indiscriminate. Patch match is non-local means with a 3 × 3 patch and 7 × 7 search window, the slowest and the best at keeping repeating texture. The strength chips jump to typical values; Compare in the preview shows what you have lost.
Which method should I start with? Smart smooth. It flattens grain in flat areas such as sky, walls and skin while leaving edges alone. Switch to Speckle if the noise is isolated dots rather than an even grain, and to Patch match if fine texture like fabric or foliage is turning to mush.
Why does the result look soft? Every denoiser trades some detail for smoothness. Lower the strength until the noise is merely reduced rather than gone, then run Sharpen Image with a small radius to bring back edge contrast. Denoise first, sharpen second; the other order amplifies the noise.
Why is Patch match slow? It compares every pixel's neighbourhood with dozens of nearby neighbourhoods and averages the ones that match, which is hundreds of operations per pixel. A 2-megapixel photo takes about five seconds and a 12-megapixel one well over half a minute, so run it on a crop or a downsized copy first. A quick blurred preview appears immediately while it runs.
Can it remove JPEG blockiness? Partly. Speckle at strength 2 or 3 hides the small blocks around edges, and Smart smooth removes the mottling in flat areas. It cannot recover detail the compression threw away.
You might also need