1. Home
  2. Blog
  3. How to fix grainy photos, and how much a free filter really helps
Guide

How to fix grainy photos, and how much a free filter really helps

Cover image for: How to fix grainy photos, and how much a free filter really helps

The short answer: grainy photos are caused by too little light reaching the sensor, so the camera boosts the signal and the boost amplifies random speckle. You fix a photo you already have by running a noise filter, and you prevent the next one by giving the camera more light. We tested a free filter on five photographs with known, added noise, so we can tell you how much it helps and where it stops helping.

Why photos look grainy

A camera sensor counts light. In a dim room there is not much light to count, so the signal is weak, and the camera compensates by raising its sensitivity (the ISO setting, or the phone's equivalent). Raising sensitivity does not add detail. It amplifies everything, including the random electrical variation that is always present, and that variation shows up as coloured speckle, especially in flat, dark areas such as a night sky, a wall or skin in shadow.

Phones make this more likely for three reasons. Their sensors are tiny, so they collect little light. Digital zoom crops the image and enlarges the noise along with it. And brightening a dark photo afterwards, in an editor, lifts the shadows and makes hidden noise visible. If a photo looked fine on the phone and grainy after you edited it, that is why.

Stop it before you take the photo

Prevention beats repair, because no filter can bring back detail that was never captured.

  • Add light. Turn on a lamp, face a window, or move away from the corner of the room. One extra light source does more than any setting.
  • Use night mode. Most modern phones have a mode that takes several frames and combines them, which averages noise away. Hold still for the full exposure.
  • Do not use digital zoom. Move closer, or crop later. Cropping afterwards gives the same result without the extra processing.
  • Clean the lens. A smudge looks like haze, and haze makes you raise exposure.
  • Tap on the subject to set exposure, so the camera does not guess too dark and then brighten the whole frame.

Fix a grainy photo you already have

You have three realistic options. Phone and desktop editors such as Lightroom and Photoshop include noise-reduction sliders, and recent versions have AI-based denoising that learns what real detail looks like; they are the best choice when you have them. Your phone's own editor often has a "noise" or "grain" setting. The third option is a free filter that runs in your browser: our reduce image noise tool, which needs no sign-up and never uploads the photo.

It uses an edge-preserving bilateral filter. Each pixel is replaced by a weighted average of nearby pixels, but a neighbour counts for less the more its colour differs from the centre pixel. Flat areas, where grain is most visible, get smoothed, while a sharp edge is mostly left alone. There are three strengths: Light, Medium and Strong.

What we measured

To know how well a filter works you need to know what the clean picture looked like, so we did not use real grainy photos. We took five clean photographs, added random brightness and colour noise at two strengths, cleaned them with the tool running in the real page in headless Chromium, saved the result as JPG at quality 90, and compared each result with the clean original. PSNR and SSIM measure closeness to the original: higher is better, and SSIM of 1.000 means identical.

Added noiseBefore cleaningLightMediumStrong
Mild: PSNR31.834.334.332.7
Mild: SSIM0.7640.8840.8960.864
Heavy: PSNR26.027.530.031.2
Heavy: SSIM0.5210.5910.7280.816

The pattern is the one you would hope for. For mild grain, Medium gave the closest result (Light tied it on PSNR), and Strong scored lower than either, because it started removing real detail along with the grain. For heavy grain, Strong was best on all five photos, and the improvement was large: SSIM went from 0.52 to 0.82. In practice that means you should pick the weakest strength that makes the grain stop bothering you, not the strongest.

Caveat: this is synthetic noise on small test photos, which lets us score the result exactly. Real camera noise is patchier and varies by colour channel, so your photos will not behave identically. The free filter also loses to AI denoisers on fine texture such as hair and fabric.

Cleaner photos are also smaller files

Grain is random detail, and random detail is the hardest thing for a JPG to store. We saved the same photos as JPG at quality 90 in three ways and compared the sizes.

VersionMild grainHeavy grain
Grainy photo, no cleaningbaselinebaseline
Cleaned with Light24% smaller9% smaller
Cleaned with Medium39% smaller26% smaller
Cleaned with Strong48% smaller43% smaller
The original clean photo30% smaller43% smaller

Part of that saving is removed grain and part is removed detail, so a smaller file is not automatically a better photo. It also explains a common surprise: a grainy night photo is often noticeably bigger than a clean daylight photo, even at the same pixel size. If you are fighting an upload limit, cleaning the grain first and then using compress to a target size gives a sharper result at the same limit than compressing the grainy file directly.

When grain is not a problem

If you will only ever view the photo small, on a phone screen or as a thumbnail, grain is mostly invisible, and shrinking the photo hides it for free. Resizing a photo to half its width averages four pixels into one, which reduces random noise on its own. In our test, halving the dimensions raised the similarity to the clean photo from a PSNR of 31.8 to 38.6 for mild grain and from 26.0 to 32.8 for heavy grain (each compared with an equally resized clean copy). The resize tool does this, and it is worth trying before the stronger filter settings if the photo does not need to be large.

Quick steps

Open the noise reduction tool, choose Medium, drop in your photo, and compare the result with the original at full size. If grain is still visible, try Strong. If faces look waxy or text looks smeared, go back to Light. Then save as JPG for the smallest file, or PNG if you plan to edit it again.

Reduce image noiseRemove grain from dark or grainy photos — free, no upload
Compress a JPGReduce JPG file size — free, no upload
Resize an imageScale to exact pixels or percent — free, no upload

Tools mentioned in this article

More from the blog

WebP vs PNG: which is smaller, and when should you use each?
WebP is usually smaller, but lossless WebP depends on the encoder. Measured sizes for six kinds of image, including a correction on our own tool.
5 min read
How to compress an image to 20, 50 or 100 KB (and what you lose)
Tested on real photos: what a 20, 50, 100, 200 or 500 KB limit does to quality and dimensions, and how to avoid a blurry result.
5 min read
We checked all 28 image conversions pixel by pixel
28 conversions, each run in headless Chromium and compared against an independent decoder. What is identical, what is lossy, and why.
4 min read
How big will your image PDF be? We measured it
Page size does not change the file size, but quality does, a lot. Measured PDF sizes from JPG and PNG sources.
4 min read
JPG vs WebP quality settings: what each level really costs in file size
Nine quality settings, two formats, five photos, measured. Where the file-size curve is gentle, and where it falls off a cliff.
4 min read
PNG vs JPG vs WebP: what we measured on photos, screenshots, text and charts
There is no best format, only a best format for each kind of image. Measured sizes and quality scores for photos, screenshots, text and charts.
4 min read
Resize or compress first? Measuring what actually shrinks an image
Image size scales with pixel count, so resizing beats lowering quality. Here are the measured numbers, and the order that works best.
4 min read
Writing a real PNG compressor in vanilla JavaScript (no WASM, no libraries)
Median-cut quantization, Floyd–Steinberg dithering, hand-built PNG chunks, and the four browser traps that cost me an evening each.
6 min read
Are online image converters safe? What actually happens to your files
Most converters upload your file to a server you know nothing about. Here's how to tell which ones do, and when it actually matters.
6 min read
JPG vs PNG vs WebP vs AVIF: which image format should you use?
Five formats, one decision. Here's what each one is actually good at, with a rule of thumb you can apply in ten seconds.
6 min read
How to combine multiple images into one PDF
Every method, on every device — plus the page-size and quality settings that decide whether your PDF is 200 KB or 20 MB.
6 min read
How to make a favicon: every size and format you actually need
Which sizes you actually need, the exact HTML to paste, and the three reasons a favicon usually fails to appear.
6 min read
How to reduce image file size without losing quality
Do these three things in the right order and a 4 MB photo becomes 80 KB with no visible difference. The order matters more than the tools.
6 min read
What is a WebP file, and how do you open one?
Why the internet keeps giving you files you can't open, how to open them anyway, and when WebP is actually the right choice.
6 min read