One grainy night photo, and how to rescue it

Last night you took a few pictures at a birthday dinner. On your phone they looked fine: bright enough, everyone recognizable, the restaurant's warm light glowing. This morning you open the best one on a laptop and look a little closer. It seems someone sprinkled fine sand all over it.

This article follows that one photo from the moment you notice the grain to the moment you decide to stop working on it.

The photo: a birthday dinner, shot on your phone

Say your photo looks like this. A busy restaurant with low lights, tables close together, a cake just arriving. You held up your phone, took one shot and put it away to get back to the party. No tripod, no flash, no second try.

To be clear from the start, this is a typical case set up so the article has something concrete to follow, not a particular photo that went through a test. But it's typical enough that you probably have one just like it.

On a small, bright phone screen, glanced at for three seconds, nothing looks wrong. Opened at full size on a laptop, three things show up at once:

• The dark areas — under the table, the corners, a dark sweater — aren't smooth. They're speckled with tiny dots of color, red, green and purple, where there should be quiet darkness.
• Faces look slightly flat, as if seen through a very thin film, even though the camera clearly found focus.
• The wall behind everyone looks sandblasted.

But the edge of the table is still crisp, and the writing on the menu is still readable. Those two small details decide everything that follows.

Grain or blur? Look before you fix anything

This is the grain branch of a bigger question. If you can't yet tell what's wrong with your photo, the guide to sorting out blur, grain and everything in between walks through the signs one at a time, and a few minutes there can save you a wasted run.

At a glance the two problems look alike, but they're opposites.

Blur is detail that has smeared. Either the camera moved when the shutter fired or the focus landed behind your subject, and the outline of everything got dragged into streaks. Zoom in and you won't see grain; you'll see mush.

Noise is a layer of speckles sitting on top of a photo that is otherwise sharp. In low light the camera has to boost its sensitivity, and the price is random dots across the image, worst in the dark areas where the real signal is weakest.

The dinner photo has a crisp table edge and a readable menu, so the lens focused where it should. What's covering the picture is grain. It's noise, not blur.

The mistake: running Deblur on a grainy photo. The photo looks bad, and "deblur" sounds like the fix for a bad photo. But deblurring looks for small details that have been smeared and pulls them back into focus, and it has no way to tell which tiny dots are real detail and which are grain; to it, both are small things to sharpen. The result can come out sharper and dirtier at the same time, with every speck in the dark areas given a crisp little edge. That's why naming the problem comes before any fix, not after.

Why night shots get grainy, and why some get smeared instead

The engineers who built Google's Night Sight explain the grain in plain terms. In their 2018 write-up they say: "For smartphone cameras, which have small lenses and sensors, a major source of noise is the natural variation of the number of photons entering the lens, called shot noise." A second source, which they call read noise, comes from random errors when the electronic charge from each pixel is turned into a number.

The classic camera explanation says the same thing in plainer words. Nikon calls noise "the digital equivalent of grain" and describes raising the ISO in a digital camera as a trade-off: more sensitivity, more noise. It also explains why small cameras struggle more than big ones: "Large pixels result in less noise than small ones. That's why digital SLRs perform much better at high ISOs than compact cameras." A phone's tiny sensor sits at the hard end of that scale.

There's a simple fix in theory, and it's the reason night modes exist. Google's write-up continues: "SNR rises with the square root of exposure time (or faster), so taking a longer exposure produces a cleaner picture." The catch is in the very next sentence: "But it's hard to hold still long enough to take a good picture in dim light, and whatever you're photographing probably won't hold still either."

That catch is the other half of the story. Night modes on today's phones gather light over a longer stretch of time. Apple says it plainly: "Depending on how dark the scene is, your iPhone might take a Night mode photo quickly, or it might take several seconds." Google's Pixel help gives the matching instruction: "Hold still for a few seconds while your device captures the photo." If your hands or the people at the table move during those seconds, the photo comes out smeared rather than grainy. That's a different problem with a different fix, and it comes up again below.

Denoise first, enlarge later

The first decision is easy now, because the hard part is done: run the Denoise mode of Deblur and denoise, not the Deblur mode. The tool runs one mode per pass. That sounds like a small annoyance, but the two modes solve opposite problems, so it makes sense to pick one on purpose.

The second decision is less obvious and matters more: don't enlarge the photo yet. Night shots from phones are often small, so "let's upscale it while we're at it" comes naturally. The reason to wait is about order, not about tools. Enlarging copies everything in the photo onto a bigger canvas, and the grain is part of what's in the photo. Clean first and you hand the enlarger a tidy picture. Enlarge first and you hand the cleanup a messier picture than the one you started with.

One fact makes this order painless: Denoise isn't an enlarger. It cleans the photo at the size it already is, so cleaning first doesn't cost you any pixels.

The mistake: enlarging first, denoising second. The photo is grainy and small, so fixing the size first feels efficient. But an enlarger can't tell grain from detail, and it builds out everything it sees. Afterward the speckles haven't gone anywhere; they've grown into small blotches. The cleanup now has more to clear, while the real detail was already diluted when the photo was enlarged.

What the first pass gives, and what it takes

Your photo is uploaded over an encrypted connection and processed automatically, and it's deleted on the schedule our Privacy Policy describes.

What you gain shows right away. Dark areas go back to being dark, without the colored dots. The wall loses its sandblasted look. The photo holds up at a larger size on screen.

What you lose is quieter, and any noise reduction pays it: some very fine detail goes away along with the grain. The weave of a sweater softens. A few stray hairs at the edge of a face disappear. Small print on the menu blurs slightly compared with the original.

That isn't a defect of one tool. It's the cost of telling grain apart from fine detail: on a photo, a speck of noise and a thin hair are both tiny marks side by side, and whatever clears one kind touches the other. Nikon's guide adds a fact that matters here: "Because all pixels collect some noise, every digital camera runs processing on every image (although with a NEF, or RAW, file that can be changed later) to minimize that noise." A phone is a digital camera too, so the photo you're fixing has usually been through one round of noise reduction already. Every pass you add stacks on top of that, which is one more reason to stop early.

Now compare the result with the original, and look at the faces first. A slightly softer tablecloth goes unnoticed. A face is where everyone spots a problem fastest, and if the faces look right, the photo is right.

Should you run it again?

The honest answer is usually no.

The temptation is easy to understand. The first pass helped a lot, but zoom in and there are still a few specks in a dark corner, so "one more pass to get it clean" sounds reasonable. The trouble is that what you lose keeps adding up while what you gain keeps shrinking: a second pass usually removes far less grain than the first, yet it still takes real detail with it.

A simple way to decide: if only a few small dark patches still show grain and the faces look good, stop. If the whole photo still looks evenly grainy, as if nothing happened, the problem is something else, and a second pass won't solve it. When a pass uses credits, each extra pass uses them again, and the cost of a pass is shown in the editor before you start it.

The mistake: denoising until the photo is "completely clean." Each pass does its job, merging small specks into smooth patches. Stack passes on top of each other and they tend to start merging specks that were real detail. Skin turns flat and waxy. Fine lines and pores vanish, and hair sticks together in clumps. The photo is clean, and it obviously went through software. A night photo is supposed to have some grain, and the little that's left is part of what makes it look like a photograph.

Want a print? Now you can enlarge it

Now the photo is clean, and enlarging it finally makes sense.

One rule is worth knowing: Upscale image builds new detail out of what it can read in the photo, so it does its best work on a photo that's already fairly clear. The cleaner the input, the more trustworthy the added detail. That's why "clean first, enlarge second" isn't a matter of taste. For how many pixels a given print size really needs, see what resolution a large print actually needs.

There's one special case. If, after denoising, the only thing that still bothers you is the faces — the scene looks fine but the faces lack sharpness, which happens a lot when people sat far from the camera — Enhance faces is built for that job and beats another pass over the whole photo. Group shots have quirks of their own, and rescuing soft faces in the back row of a group shot covers them.

Before you order the print, open the downloaded file and look at the corners. A download may include a small vipix.vn mark; whether it does comes down to the tool and to whether you were signed in when you ran it.

When a night photo can't be saved

There's a group of photos nothing in this article will help, and it's better to say so than to hide it: shots that are too dark, heavily grainy and shaken all at once, with the subject out of focus. There isn't enough information left for any tool to rebuild. Clear the grain and you uncover mush; sharpen the mush and the grain comes back. What you get isn't a fixed photo, just a photo broken in a different way.

Light is the root of it. When Google introduced Night Sight in 2018, it said: "Smartphone cameras that take a single picture begin to struggle at 30 lux. Phones that capture and merge several pictures (as HDR+ does) can do well down to 3 lux" — and further down, "Below 0.3 lux, autofocus begins to fail. If you can't find your keys on the floor, your smartphone can't focus either." Those are 2018 figures for 2018 phones, not a spec for yours. The principle hasn't changed, though: below some level of light, the camera can't find an edge to focus on, and without an edge there's nothing left to rescue.

The quick test is to open the original on a computer at full size and look for one spot, any spot, with a crisp edge: the rim of a glass, the frame of someone's glasses, the letters on a sign. If you can't find a single one, the photo has lost the ground you'd need to stand on, and every pass after that only trades one kind of damage for another.

The mistake: expecting Denoise to fix camera shake. The photo is dark and smeared, and "it's a night photo, so denoise it" seems logical. But shake and grain are different kinds of damage. Shake drags detail into streaks, so the information has been mixed together and is no longer where it belongs. Denoise only clears the specks on top; it has nothing that could move the dragged detail back. The result is cleaner and just as smeared, sometimes worse, because the grain that partly hid the smear is gone. For a photo that's both smeared and grainy, the order is Deblur first, then Denoise. And if you're not sure whether the smear came from your hand or from someone moving, whether it's camera shake or a moving subject shows how to tell.

If the scene can still be shot again, retaking it is the shortest route, even if that sounds lazy.

Next time: a cleaner night shot

Most of the rescue above can be avoided in the seconds around the shot.

• Let night mode work. Apple's support page says: "You can use Night mode with iPhone 11 and later." On a Pixel, Google notes that Night Sight turns on by itself when you shoot in low light in Photo or Portrait mode, with one exception worth remembering: "When the camera flash is on, the Automatic Night Sight doesn't work."
• Hold still, and brace the phone. Apple's advice: "For the best results, hold your iPhone steady until the capture completes. Try placing your iPhone on a solid and secure surface, or use a tripod to increase stability and control of the exposure clarity." A table edge or a wall works when there's no tripod.
• Ask everyone to hold the pose until the capture finishes. Night modes can need a few seconds, and anyone who moves during them comes out smeared.
• Move toward the light. The more light reaches the sensor, the less the camera has to boost it and the less grain you'll have to clean later.

And if this one can't be reshot, because that dinner only happens once, take the route this photo took. Name the problem first. Clean before you enlarge, and enlarge only if you need a bigger copy. Then stop one pass earlier than feels natural.

Frequently asked questions

Why are my night photos so grainy?

In low light the camera boosts its sensitivity, and noise rises with it. Google's camera engineers point to shot noise, the natural variation in how much light reaches the sensor, plus read noise from the electronics, and small phone sensors make both harder to avoid.

How can I tell whether my photo is grainy or blurry?

Look for one crisp edge, such as the rim of a table or a line of text. If edges are crisp but the dark areas are speckled, it's grain. If edges are smeared into streaks, it's blur.

Should I denoise or upscale first?

Denoise first. Enlarging builds out the grain along with the detail, while cleaning first gives the upscaler a cleaner photo to work from.

Why does my photo look waxy after noise reduction?

Too much noise reduction removes fine detail along with the grain. Digital cameras, phones included, already reduce noise in every image, so an extra pass stacks on top of that. Repeated passes tend to flatten skin and hair.

My Night mode photo came out smeared. Can denoising fix it?

No. Night mode can take several seconds, and movement during that time smears the photo, while denoising only removes grain. Use the Deblur mode for the smear, and if the photo is also grainy, run Deblur first and Denoise after it.