How to remove noise from a night photo

Where grainy low-light photos come from

If you want to remove noise from a night photo, it helps to know what you are removing. Every camera sensor turns light into an electrical signal, and every sensor also adds a small amount of random variation to that signal. In daylight the signal is so strong that the variation disappears under it. At night, or in a dim living room, the signal is weak and the variation is suddenly a visible part of the picture. That is noise.

Three things make it worse, and they usually arrive together:

• Little light. Fewer photons reach each pixel, so the ratio between real signal and random error drops. A street at 10 pm or a restaurant lit by candles is a hard scene for any camera.
• A small sensor. Phone sensors are a fraction of the size of those in dedicated cameras, and each pixel on them is tiny. Tiny pixels collect less light, so they show noise sooner. That is why the same scene can look clean on a large camera and gritty on a phone.
• High ISO. ISO is amplification. When the camera raises it to make a dark scene look bright, it amplifies the noise along with the image. A shot taken at ISO 3200 or 6400 will almost always be grainier than one at ISO 100, because the camera is turning up the volume on a weak signal.

Not all noise looks the same, and the difference matters for how you clean it. Luminance noise is variation in brightness: fine, speckled grain that looks a bit like film texture. In moderate amounts it can even be pleasant, because the eye reads it as detail. Colour noise (also called chroma noise) is variation in colour: blotches of green and magenta scattered over areas that should be a smooth, even tone, such as a dark sky, a wall or the shadow side of a face. Colour noise is the one that makes a photo look cheap and broken, and it is also the easier one to remove without losing real detail, because genuine detail rarely lives in those random colour specks.

One more thing to know: phones already apply their own noise reduction before you ever see the file. When that processing is heavy, you get smeared, watercolour-like areas instead of grain. Those smears cannot be undone afterwards, which is another reason to get the shot right in the first place.

How to avoid high ISO noise when you shoot

The cleanest noise reduction is the one you do not need. A few habits at the moment of taking the photo will give any later cleaning much better material to work with.

1. Steady the phone. Lean it against a wall, a railing, a glass on the table or the back of a chair. A steady camera can use a longer exposure instead of a higher ISO, and a longer exposure gathers more light, which means less noise. A small tripod costs little and makes a visible difference for city and landscape shots at night.
2. Use night mode. Most phones have one, and it usually switches on by itself when the scene is dark. It takes several frames over one to a few seconds and merges them, which averages out a large part of the random noise. Hold still until the counter finishes; moving halfway through gives you ghosting or blur.
3. Do not push the ISO up by hand. In manual or pro mode it is tempting to set ISO 3200 to get a bright preview. It is usually better to leave ISO on auto, or as low as the scene allows, and let the exposure time do the work if the phone is steady. A slightly dark photo at a lower ISO often cleans up better than a bright one at a very high ISO.
4. Add light where you can. Switch on another lamp, move the person closer to a window or a shop front, turn towards the light source. Every bit of extra light improves the signal more than any software can later.
5. Do not crop heavily afterwards. Cropping a small part of the frame and blowing it up is the same as enlarging the noise. If you crop a night photo to a quarter of its area and view it at the same size, every speck of grain is twice as wide and twice as tall. Get closer when you shoot, or use the phone's optical telephoto lens if it has one, rather than relying on digital zoom or a tight crop.

If you shoot in RAW, the file keeps more information in the shadows, but it also skips the phone's own noise reduction, so a RAW night shot will look noisier straight out of the camera. That is normal: the cleaning is simply left to you.

How far to go when you clean a dark photo

Noise reduction always works by guessing which small variations are random and which are real detail, and then smoothing the random ones. The trouble is that real detail and noise overlap at the same tiny scale. Skin pores, the weave of a jumper, the grain of wood, fine hair, the texture of a brick wall and the stars in a night sky all look, to an algorithm, a lot like noise.

Push the cleaning too far and you get the familiar overcooked look:

• faces turn smooth and waxy, like a plastic doll, with no pores or fine lines left;
• clothes lose their fabric texture and become flat patches of colour;
• faint stars and distant lights vanish from the sky along with the grain;
• foliage and grass merge into a soft green blur;
• edges stay sharp while everything between them goes flat, which looks painted rather than photographed.

A photo with a little fine grain left in it looks like a photograph taken at night. A photo with no grain at all and no texture looks like an illustration. Most people prefer the first, even if they cannot say why.

That is why a medium setting is usually the right starting point. The aim is to get rid of the colour blotches completely and to tame the luminance grain until it stops distracting, not to erase it. In practice that often means somewhere around the middle of the intensity range, and lower for portraits where skin texture matters.

The other rule is to compare at 100 %. On a phone screen, or in a small preview, almost any amount of noise reduction looks good, because the image is shown much smaller than its real size and the loss of texture is invisible. Zoom in to see the pixels one to one, look at a face, an eye, a patch of fabric and a dark area of the background, and compare before and after. If the before has detail you can see and the after does not, back off.

Finally, accept that cleaning has limits. Noise reduction can remove random variation; it cannot recreate detail that the sensor never recorded. A very dark, very noisy photo will come out cleaner, but not as sharp as one taken in good light. And if the photo is also blurred by movement, that is a different problem: unblur a photo is a separate step, best done after the noise has gone.

Denoise first, then upscale or restore

Noise reduction is rarely the only thing you want to do to a night photo. You might want it bigger for printing, or you might be rescuing an old, grainy picture with faces in it. The order of the steps changes the result a great deal.

Clean before you enlarge. An upscaler is built to find structure in an image and draw it more finely at a higher resolution. If the image is full of grain, the upscaler treats the grain as structure, sharpens it and spreads it over four or sixteen times as many pixels. You end up with bigger, crisper noise. Run the noise reduction on the original file first, check it at 100 %, and only then upscale the image. If the photo is going on a wall, the guide to upscaling an image for large-format printing explains how much resolution you really need for the viewing distance.

Clean before you restore faces. Face restoration models look for eyes, mouth and skin and redraw them with fine detail. Strong colour noise around a face confuses that detection and can push the model to invent features that are not there. A cleaner input gives a result that looks more like the actual person. When you restore old photos that were shot on fast film or scanned from small prints, a light, separate pass of noise reduction first, at about half strength, is often the better order. Our guide on how to restore old photos with AI covers the rest of that workflow.

Compress last. Compression to JPEG or WebP also struggles with noise: random grain is hard to compress, so a noisy photo produces a bigger file, and the compression artefacts mix with the grain into an ugly mess. Clean the image, do any enlarging, and then compress the image for the web or for sending by message as the final step.

A sensible order for a typical night shot is therefore: denoise, then unblur if needed, then upscale or restore, then compress. Each step works on the cleanest possible version of the previous one.

How to remove noise from a night photo in Airpic, step by step

Airpic works in the browser, with no account and no subscription. Processing and looking at the result is free; the full-resolution download follows the rules shown on the tool page. Files are deleted one hour after download and after one day at the latest, they are never used for training, and the GPUs that process them are in the European Union.

1. Open the tool and upload your photo. It accepts JPG, PNG (including transparency and 16-bit), WebP, TIFF, HEIC and AVIF, so you can drop in an iPhone picture as it is. The EXIF orientation is applied and the colour profile is kept.
2. Use the original file. Upload the photo straight from the camera or phone, not a version that has been sent through a messaging app, cropped or already compressed. Every extra round of compression bakes more artefacts into the grain.
3. Set the intensity slider. Start at a medium value. Go lower for portraits and for skies with stars; go higher for flat surfaces such as walls, dark skies without stars and night interiors where texture matters less.
4. Process and compare. Use the before-and-after comparison and switch to the 100 % view. Check a face, a fabric, a dark area and any fine lights. If skin looks waxy or stars have disappeared, lower the intensity and run it again; if colour blotches remain in the shadows, raise it a little.
5. Choose the output format and download. PNG keeps everything; JPEG is lighter and fine for sharing. If you plan to enlarge or restore the photo, download the cleaned version now and use it as the input for the next tool.

Ready to try it on your own shot? Open the photo denoiser to remove noise from a night photo, start with a medium intensity and compare at 100 % before you download.

FAQ

Keep reading