Denoise Image

Photos taken in dim light, indoors without flash or at night, often show noise: grainy speckles, especially in shadows and flat areas like skies and walls. Scans of old prints come with dust and tiny specks. This tool offers two methods for these two different problems, with controls to balance smoothness against detail.

Where noise comes from

A camera sensor collects light, and in low light it amplifies a weak signal. The amplification, set by the ISO value, also amplifies random variation, which appears as grain and colored blotches. Small phone sensors show it sooner than large camera sensors. Scans have a different problem: dust, scratches and paper texture create isolated specks rather than an even grain.

Smooth, keep edges

The default method uses a guided filter, a well-known edge-preserving technique. It looks at each small neighborhood: where the image is flat, the pixels are averaged and noise disappears; where there is a real edge or texture, the variation is much larger than noise and the pixels are kept. Faces stay sharp at the eyes while skin and backgrounds become clean. The filter runs in strips, so even large photos process without running out of memory.

Median for specks

The median method replaces each pixel with the middle value of its neighbors. A single bright or dark speck is an outlier and simply vanishes, while edges stay in place. It is the classic cure for dust on scans, salt-and-pepper noise, dead pixels and JPEG mosquito noise around text. On ordinary photo grain it can look painterly, so use it for specks rather than general smoothing.

Radius

Radius is the size of the neighborhood. Fine grain needs a radius of 1 or 2. Coarser blotchy noise, typical of phone night photos, may need 3 or 4. Larger radii remove more noise but also more fine detail.

Edge protection

In smooth mode, edge protection decides how much local variation counts as a real edge. Low values protect only strong edges and smooth more aggressively. High values protect more texture, keeping hair, fabric and foliage detailed but leaving some noise. Start at the default of 24 and adjust while comparing.

Strength

Strength blends the denoised result with the original. Full strength gives the cleanest result; 50 to 80 percent keeps a little natural grain, which usually looks more photographic than perfectly smooth surfaces.

Order of editing

Denoise first, then adjust brightness and colors, and sharpen last. Sharpening a noisy image amplifies the noise; denoising after sharpening blurs the sharpened edges. Brightening shadows also reveals noise, so check again after exposure changes.

Limits

No filter can tell noise from very fine detail with certainty. Extremely noisy images will lose some texture. For scans with large scratches or tears, a retouching brush in an editor works better.

Phone night photos

Modern phones already denoise their night shots heavily, which often leaves smeary patches rather than grain. Running more smoothing on those rarely helps. The tool is most useful for photos from older phones, compact cameras at high ISO, screenshots of video frames and scans, where the noise is still visible as grain or specks. Compare at full size before deciding.

Related tools

After denoising, apply Sharpen Image gently. To check whether shadows were underexposed, look at the Image Histogram. For softening backgrounds on purpose, use Blur Image.