Clean up grainy low-light and high-ISO photos in your browser. Choose a non-local means, bilateral or median filter, set the strength, keep fine detail and compare the result with the original using the before/after slider.
Last updated:
◂ ▸
OriginalDenoised
Processing…
Filter
Tip: noise is easiest to judge at 1:1 zoom. Keep a little grain for a natural look.
Advertisement
Advertisement
How to use the Image Noise Remover
Add a photo: drop it on the box, paste it or click Choose photo. A noisy sample photo is loaded so you can try the filters right away.
Pick a filter: Non-local means for most photos, Bilateral for a fast smooth result, Median for specks and salt-and-pepper dots.
Set the strength for grain and the Color noise slider for colored blotches, then raise Detail preservation if edges look too soft.
Compare: drag the divider on the image, or tick Zoom 1:1 to inspect real pixels.
Download the cleaned photo as PNG (lossless) or JPG.
How the three noise filters work
The photo is first split into brightness (Y) and two color channels (Cb, Cr) with the standard JPEG / ITU-R BT.601 formula Y = 0.299 R + 0.587 G + 0.114 B. Grain is removed from Y with the filter you choose; color blotches are smoothed separately in Cb and Cr, because the eye is far less sensitive to fine color detail than to fine brightness detail.
Filter
How it works
Best for
Speed
Non-local means
Each pixel becomes a weighted average of pixels in a 7×7 to 9×9 search window whose surrounding 5×5 patch looks similar. Weight = exp(−max(d² − 2σ², 0) / h²), where d² is the mean squared patch difference and σ the estimated noise.
Most photos: high ISO, low light, phone night shots
Slowest
Bilateral
Averages neighbors weighted by both distance and brightness difference, so pixels across a strong edge barely mix.
Smooth skies, skin, quick cleanup
Medium
Median
Replaces each pixel by the median of a 3×3 to 7×7 window, which ignores outliers completely.
Salt-and-pepper dots, hot pixels, scanned specks
Fastest
Detail preservation blends a share of the original back in, more on edges (found with a Sobel filter on the cleaned image) than in flat areas, so textures such as hair, fabric and foliage do not turn to wax.
Tips for cleaner photos
Denoise before you sharpen. Sharpening first makes grain stronger and harder to remove.
Treat color noise more strongly than grain. Colored blotches look digital and ugly; fine monochrome grain often looks like film and can stay.
Check at 1:1. A photo can look clean when scaled down while faces look plastic at full size. Tick Zoom 1:1 and compare both halves.
Save as PNG if you will edit further. JPG compression adds its own artifacts; use JPG only for the final copy.
Prevent noise at the source. More light, a lower ISO, a tripod with a longer exposure, or slightly brighter exposure (then darkening later) all give less noise than any filter can remove.
Frequently asked questions
How do I remove noise from a photo for free?
Open the photo in this tool, keep the Non-local means filter, adjust Strength until the grain disappears and press Download. Everything runs in your browser, with no sign-up, no watermark and no upload.
Which denoising filter should I use?
Non-local means gives the cleanest result on most photos because it averages pixels whose surroundings look alike, so edges and textures survive. Bilateral is faster and gives a smooth, slightly painterly look. Median is the best choice for isolated specks, dead pixels and salt-and-pepper noise.
What is the difference between luminance and color noise?
Luminance noise is the gray, film-like grain that changes pixel brightness. Color (chroma) noise shows up as red, green and purple blotches, especially in shadows. The tool converts the image to YCbCr and treats them separately: Strength works on brightness, Color noise on the two color channels.
Why does my photo look plastic or blurry?
Too much strength removes real texture together with noise. Lower Strength, raise Detail preservation, or leave a little grain: a slightly grainy photo usually looks more natural than a waxy one.
What does the noise estimate mean?
It is the estimated standard deviation of the noise in brightness levels (0 to 255), calculated with Immerkær's fast method from a Laplacian-like filter. Lower is cleaner. Very detailed textures can raise the number a little, so use it as a guide, not an exact measurement.
Is my photo uploaded to a server?
No. The filters run in a Web Worker inside your browser using the Canvas API. Your photo never leaves your device.
Is there a size limit?
Photos larger than 4000 pixels on the longest side are scaled down to 4000 px so processing stays fast. Non-local means on a 12-megapixel photo can take several seconds; a progress bar shows how far it is.
Are all the tools on FreeToolCabin tested for quality and accuracy?
Yes! Every tool on our site undergoes automated Q/A testing in real browser environments to ensure high precision, correct output, and zero errors. If you happen to spot a bug or have a suggestion, please contact us at [email protected] and our team will review it right away.