Image Denoiser
Get rid of white and black specks, dust spots, dead pixels and salt-and-pepper noise. The denoiser uses an adaptive median filter: each pixel is compared with its neighbours and replaced only when it stands out as an outlier, so clean areas, fine lines and text remain exactly as they were. Choose the size of the neighbourhood, the sensitivity and the number of passes, and see how many pixels were corrected.
- Files stay on your device
- No sign-up
- Free to use
For fine grain from high ISO, the Image Noise Reducer works better; this tool targets isolated specks.
Despeckle
A median filter replaces isolated outliers – white and black specks, dust, salt-and-pepper noise, stuck pixels – with the typical value of their neighbours. The threshold leaves normal pixels untouched, so detail and edges stay sharp.
Output
How to use Image Denoiser
- Drop scans or photos onto the page.
- Choose the neighbourhood size for the size of the specks.
- Set the threshold so only real specks are changed.
- Download the cleaned images.
Image Denoiser features
Adaptive median
Only outliers are replaced.
Three sizes
3×3, 5×5 and 7×7 neighbourhoods.
Threshold
Protects detail, text and edges.
Counter
Shows how many pixels were corrected.
Batch
Apply the same settings to up to 50 images.
Private
Processed in your browser; nothing is uploaded.
When to use Image Denoiser
- Cleaning dust from scanned slides, negatives and prints.
- Removing specks from scanned documents.
- Fixing stuck or hot pixels in camera photos.
- Removing salt-and-pepper noise from old digital images.
Image Denoiser FAQ
What is salt-and-pepper noise?
Isolated pixels that are much brighter or darker than their surroundings, like grains of salt and pepper. It comes from transmission errors, faulty sensor pixels, dust and scanner defects.
Why a median and not a blur?
A blur spreads a speck over its neighbours and leaves a smudge. The median picks the typical value of the neighbourhood, so a single outlier disappears completely while edges stay sharp.
Which neighbourhood size should I choose?
The smallest that removes the specks: 3×3 for single pixels, 5×5 for specks of two or three pixels, 7×7 for larger dust. Larger sizes can round off fine details.
What does the threshold do?
A pixel is only changed when it differs from the median by more than the threshold. Higher values touch fewer pixels; lower values clean more aggressively.
Does it remove film grain?
Not well, because grain is not made of isolated outliers. Use the Image Noise Reducer for grain.
Are my photos uploaded?
No. The images are processed in your browser and never leave your device.
Median filtering for clean images
The median filter is one of the oldest and most effective tools against impulse noise. For every pixel it collects the values of the surrounding pixels, sorts them and takes the middle value. A single speck, however bright, ends up at one end of the sorted list and never becomes the result. Edges survive because the median of a neighbourhood on an edge is still a value from one side or the other, not a mix.
A plain median filter applied everywhere has a cost: it rounds corners, removes thin lines and flattens fine texture. The adaptive version used here first checks whether a pixel is an outlier. Only pixels that differ from their neighbourhood median by more than the threshold are replaced; all others keep their original value. On a typical scan, fewer than one percent of the pixels change.
Scanned photographs and film are the classic use. Dust on the glass or the film appears as white or black spots, and fibres as short lines. The 5×5 and 7×7 settings deal with larger particles; several passes catch clusters of specks that protect each other in a single pass.
For documents, the denoiser removes the scattered dots that cheap scanners and fax machines add around text, which also improves OCR results. Use a moderate threshold so that the dots of letters such as i and j are kept.
The counter under the preview tells you how many pixels were corrected, which makes it easy to find the gentlest setting that does the job. All processing happens in your browser.