Blaming the ISO number when the real problem is that too little light reached the sensor.
Editing and retouching
Image noise
Also called grain, photo noise or iso noise. Here is what it means, when it changes what you export, and what people get wrong.
Editing and retouchingRuns on your deviceUpdated 10 August 2026
The short answer
Image noise is the random speckle in a photo that does not correspond to anything in the scene. It comes from the sensor and its electronics, and appears as brightness flecks (luminance noise) and coloured blotches (chroma noise). Dim light, high ISO settings and underexposed shadows all make it more visible.
Updated
The short version
Image noise at a glance
| Quick fact | Detail |
|---|---|
| Two kinds | Luminance (grain) and chroma (colour blotches) |
| Main physical cause | Shot noise, the randomness of arriving light |
| Worst in | Shadows, high ISO and long exposures |
| Effect on file size | Increases it, because noise compresses badly |
| Cheap to remove | Chroma noise |
| Expensive to remove | Luminance noise, it takes texture with it |
Slide the table sideways to see every column.
Where does noise actually come from?
Most of it is the light itself. Photons arrive at random intervals, so two neighbouring sensor wells collecting the same average brightness will not collect exactly the same count. That randomness is called shot noise, and it is proportional to how little light you gathered. Collect plenty of light and the variation is tiny next to the signal. Collect very little and the variation is a large fraction of it.
On top of that sits read noise from the electronics that convert charge to numbers, and thermal noise that builds up during long exposures as the sensor warms. This is why the usual advice about ISO is slightly misleading: raising ISO does not manufacture noise, it amplifies a weak signal that already had a poor ratio of image to randomness. Larger sensors and larger pixels look cleaner mainly because they catch more light in the first place.
What is the difference between luminance and chroma noise?
Luminance noise varies brightness. It looks like film grain, it follows edges reasonably politely, and in moderation people accept it or even like it. Chroma noise varies colour, showing up as red, green and magenta mottling in shadows and flat areas. It always looks wrong, and it is what makes a dark photo look cheap rather than atmospheric.
The practical consequence is that they deserve different treatment. Chroma noise can be reduced hard with very little visible cost, because real colour information in a photo changes slowly across an image. Luminance noise reduction is expensive, because the fine variations it removes are the same scale as fabric weave, wood grain and skin texture. Push it too far and a jumper turns into a plastic shell.
Why does noise make your files bigger?
Compression works by predicting what comes next and only recording the difference. Noise is by definition unpredictable, so it defeats that entirely. In a JPEG or WebP encoder the high-frequency coefficients that would normally round to zero in a smooth area stay stubbornly non-zero, and every one of them costs bits.
The result is that a noisy photo can be substantially larger than a clean one at the same quality setting, or, if you have pinned the file size, the encoder spends its budget on the grain and gives you less for the product. That matters where platforms cap upload size: Etsy warns that files over 1 MB may fail to upload, and Walmart Marketplace allows a maximum of 1 MB per file. Reducing noise before you compress is one of the few adjustments that improves both file size and appearance at once.
How does noise affect cut-outs and upscaling?
Background removal has to decide, pixel by pixel, what is subject and what is not. In a noisy shadow region the boundary is genuinely ambiguous, so the alpha edge comes back ragged, and fine detail such as hair or fabric fringe picks up speckle that reads as dirt against a white background.
Upscaling has the opposite problem. A model asked to add detail sees grain and treats it as texture worth reinforcing, so a mild speckle at the original size can become obvious blotching at double the dimensions. The order that works is to clean the noise first, then upscale, then cut out, then sharpen last of all.
What people get wrong about image noise
Each one is a real failure mode, not a style preference.
Sharpening a noisy image, which promotes every grain speck into a hard white dot.
Compressing a noisy photo harder to squeeze under an upload limit, which sacrifices product detail while the encoder still spends bits on grain.
Running noise reduction after upscaling instead of before, by which point the grain has already been magnified into blotches.
Questions people ask
Image noise, answered
The follow-up questions people search for once they have the definition.
Does high ISO cause noise?
Not directly. ISO amplifies the signal the sensor already captured, noise included. You reach for high ISO when light is short, and short light is the actual cause. The same ISO setting in bright conditions produces a much cleaner image than it does in a dim room.
Why is my photo noisy even at low ISO?
Usually because it was underexposed and then brightened in editing, which amplifies the noise exactly as raising ISO would have. Small sensors and long exposures that warm the sensor also contribute. Check the darkest areas first, since shadows show noise long before midtones do.
Can noise reduction ruin a photo?
Yes, if you push luminance reduction hard. It cannot tell grain from fine texture, so fabric weave, hair and skin flatten into a waxy surface. Reduce chroma noise generously, since colour detail is naturally coarse, and go carefully with luminance.
Does noise affect background removal?
It does. A noisy edge is an ambiguous edge, so masks come back ragged and speckle clings to hair and fringe detail, which shows plainly once the subject sits on white. Cleaning the noise before cutting out generally gives a tidier alpha edge.
Same cluster
More from the glossary
Neighbouring entries, so every term in the set is one hop from every other.
Image segmentation
Image segmentation is the task of labelling every pixel in a photo with what it belongs to, rather than drawing a box around it.
Image metadata
Image metadata is the information stored inside an image file about the image rather than in the pixels: camera settings, dates, location, captions, credit, keywords and colour profile.
Inpainting
Inpainting fills a selected region of a photo with content that matches its surroundings, so whatever was there looks as though it never existed.
From the studio
Tools that do this
Each one runs on your own device, so the photo stays where it is.
Sources
Where the facts on this page come from. Every link opens the specification, standard or documentation the claim was read at.
Keep reading
Elsewhere on Photomop
Platform image rules
Image size, file type and background rules for 22 selling platforms, each one sourced.
Read itConvert image
JPG, PNG, WebP and AVIF, with an honest answer about HEIC and alpha.
Read itAll the tools
The whole toolbox in three families, with what each one needs before it can run.
Read itOn-device privacy
The pipeline in five steps, what leaves your device, and how to check it yourself.
Read it
Now do it to a photo
Photomop is a photo studio that runs on your own device. Resize, crop, compress, convert, batch edit, remove a background and change one all work in the browser tab you are reading this in, at full resolution and with no watermark. The on-device page shows you how to check that the photo stays put.