A QR code extractor reads a QR code straight out of a picture - a screenshot, a photo, a scanned flyer, a saved graphic - and decodes the link or text stored inside it, without you having to point a phone camera at a screen. Drop an image into the tool above and EZQR finds the code, shows you exactly what it points to, and gives you back a clean, print-ready copy.

How to Extract a QR Code From an Image

  1. Add your image. Click to choose a file, drag and drop it, paste a screenshot, or tap Use camera to take a photo. PNG, JPG, GIF, WEBP, and BMP all work.
  2. EZQR scans and decodes the QR code. The tool reads the QR code out of the image and shows you the exact link or text encoded in it. Copy it with one click.
  3. Download or customize the result. Get a fresh, high-resolution QR code as PNG, JPG, or SVG, or add a logo, color, or shape. Print it, resize it, or reuse it anywhere.

Everything You Can Do With the Extractor

Extract a QR code from a photo

Took a picture of a poster, sign, or business card? Upload it and pull the QR code straight out of the photo.

Decode a QR code from an image

See exactly what a QR code points to before you scan it - the full link or text, read straight from the file.

Copy a QR code out of a screenshot

Paste a screenshot or drag in a saved graphic, and copy the decoded link or download the code as a fresh file.

Scan a QR code from an image, no phone needed

Read a QR code that's already on your screen - no second device, no pointing a camera at a monitor.

Get a clean, high-resolution copy

The extracted code is regenerated sharp and watermark-free, so you can reprint it at any size without quality loss.

Private by design

Decoding runs entirely in your browser. Your image isn't uploaded anywhere - it stays on your device.

When People Use a QR Code Extractor

Recover a QR Code From a Bad Photo

Most QR code images in the wild aren't clean screenshots. They're a photo of a sign taken from the side, a code shrunk down in a scanned PDF, a picture with glare across it, or a printed code that's been creased and scuffed. The decoder here is built for exactly that: it reads codes that are tilted, rotated, small in frame, low-contrast, or hit by glare, and if the first read comes up empty it automatically retries with contrast adjustment to pull out faint codes - then, for the hardest images, tries harder with a second, stronger decoder that reads blurred, damaged, and tiny codes most apps give up on. Where your browser exposes your phone's own scanner, that is consulted first, so the result matches what your camera app would see. We test the whole chain against more than 500 real-world photos of QR codes - blurred, shadowed, glared, creased, and shot from bad angles. QR codes also carry error correction of their own, so one that's partly damaged or covered can still decode.

That's what makes this useful for restoring a code you'd otherwise have lost. Once the link is recovered, EZQR regenerates the QR code from scratch at full resolution - so a blurry photo or a low-resolution graphic comes back as a crisp file you can reprint at any size. You're not upscaling the old image; you're getting a brand-new code that encodes the same destination. So if you came looking for a QR code cleaner or a QR code upscaler - or just a way to grab or duplicate a QR code from an image - this does the job better than either, because the result is sharp at any size rather than an enlarged copy of the blur.

It also reads codes that aren't flat or printed at all. People have brought us a deformed QR code tattoo that no phone would read and a worn, scratched QR code on the back of a prepaid Visa gift card that a store scanner kept rejecting - and the extractor pulled the contents out of both from a photo. It won't work on every wreck, but a gift card QR code that won't scan or a tattoo that's warped with the skin is worth one try here first. More on these cases on the online QR code scanner page.

QR Code Repair: How a Damaged Code Gets Reconstructed

If you are trying to repair a damaged QR code - torn, scratched, covered by a sticker, or only partly in the photo - this is the technical detail behind what the extractor does. It is automatic: you upload the image, and the decoder works through these steps until one of them produces a read it can check.

Missing or damaged finder patterns

The three big corner squares are how a normal decoder finds a QR code at all; lose one and most readers report "no code found" even when the data area is intact. EZQR's locator does not insist on all three. It can rebuild the grid from two finder patterns, or from one finder plus the timing pattern and the alignment pattern, then sample the modules through that reconstructed geometry. A mirrored code - printed backwards, or photographed through glass - is detected from its format bits and read the right way round.

A QR code cut off at the edge of the photo

When a code runs off the edge of the frame, a border, or a letterbox bar, the missing strip is not guessed at. Those squares are marked as unknown and handed to error correction to solve, with the cut edge located from the finders that are still visible. That is often enough to reconstruct a QR code from a partial image, as long as the missing strip is within what the code's error correction level can cover.

Going past the Reed-Solomon limit

A standard decoder treats every bad codeword as an error, and an unknown error costs two parity codewords to fix. EZQR tracks which squares it genuinely could not read and treats them as erasures, which cost one - and it solves them at the bit level, so a half-legible codeword only spends parity on the bits that are actually missing. Damage that is spread across the code, which defeats byte-level Reed-Solomon decoding fastest, is the case where this helps most. For codes damaged beyond what their parity can prove, the decoder can still fit the remaining data to the shape a web address takes; when it does, the result is clearly marked as unconfirmed, so you check it before you open it.

See exactly what was repaired

Every repaired read comes with a repair view over your photo: blue squares were hidden or unreadable and filled in, orange squares were read as the wrong colour and corrected, and a toggle flips back to the original image. Then download a freshly rebuilt copy as PNG, JPG, or SVG - a new code with the same contents, not an upscaled photo.

When ZXing or your phone fails to decode it

If you are a developer whose library returns nothing, this is a quick second opinion. The page runs a chain of decoders in a background worker in your browser: the browser's built-in barcode detector where one exists, then ZXing, an adaptive-threshold retry, WeChat's open-source QR engine, BoofCV, and finally EZQR's own erasure decoder for damaged codes. If none of them read it, it tells you whether a real code was found but could not be recovered, or whether there is no QR code in the image at all.

What's Inside a QR Code

A QR code isn't a picture of your link. It's a grid of light and dark squares called modules, and every one of them has a job. The map below is a real version 2 code (25 × 25 modules) at error-correction level M, drawn from the same layout tables our decoder uses to read and repair codes.

001122334455667788991010111112121313141415151616171718181919202021212222232324240123456141312111098776543210891011121314D1D2D3D4D5D6D7D8D9D10D11D12D13D14D15D16D17D18D19D20D21D22D23D24D25D26D27D28E1E2E3E4E5E6E7E8E9E10E11E12E13E14E15E16
Blue: 28 data codewords. Pink: 16 error-correction codewords. Orange: format information. Dark grey: the fixed patterns a scanner uses to find and line up the code. Each thin arrow runs through one codeword's eight bits in reading order.
  • Codewords are 8 bits. Your content is turned into a stream of bytes, and each byte fills eight modules.
  • They snake through the grid. Placement starts in the bottom-right corner and climbs a column two modules wide, then turns and comes back down the next one, working right to left. It steps around the finder, timing, and alignment patterns, which is why some codewords bend into odd shapes or split in two.
  • Error correction comes after the data. The code adds Reed-Solomon error-correction codewords calculated from the data. They're what let a scratched, smudged, or logo-covered code still decode. Bigger codes split the data into blocks and interleave them, so one damaged patch costs several blocks a little instead of one block a lot.
  • Leftover modules stay light. When the codewords don't fill the grid exactly, the spare modules are set light - 7 of them at this size.
  • A mask breaks up patterns. One of eight mask patterns flips the data modules in a fixed rhythm, so the code never ends up with big solid blocks or shapes that look like a finder pattern. The scanner undoes the mask before reading.

The 15 Format Bits

Beside the finder patterns runs a strip of format information, stored twice so one copy can be damaged. It's the first thing a scanner reads: 2 bits give the error-correction level and 3 bits say which mask was applied. The other 10 are BCH check bits, enough to fix up to 3 wrong bits, and the whole strip is mixed with a fixed pattern so it's never all light. One module beside the bottom-left finder is always dark.

14113012111110091817061504031201101EC levelMaskBCH check bits (fix up to 3 errors)
The format bits of the code above: level M, mask 3.

All that redundancy is why the extractor can often rebuild a code that's partly missing. Save or print the QR code layout infographic for the full map with example codeword shapes.

Check Where a QR Code Goes Before You Scan It

A QR code gives no hint of its destination just by looking at it, which is why "what does this QR code link to?" is worth answering before you point a camera at an unfamiliar one. Drop the image in above and EZQR shows you the full URL as plain, readable text - nothing is opened and nothing is visited. You read the destination first and decide from there. It's the same check worth running on your own printed material, to confirm the codes on a flyer, menu, or sign really do point where they were meant to.

Identify a QR Code With a Logo in the Middle

A logo sitting in the middle of a QR code tells you nothing reliable about where it goes - anyone can put any mark there, and the logo is not part of the data. To identify a QR code with a logo, decode it: drop the picture in above and read the domain in the URL that comes back. That is the only part an attacker can't dress up.

The logo doesn't get in the way of reading it, either. QR codes carry error correction, so a mark covering the middle - a brand logo, a coloured badge, an icon - still decodes normally here, the same as a plain code.

Get a Phone Number, Address, or Wi-Fi Details Out of a Code

Not every QR code holds a link. Plenty encode a phone number, an email address, a contact card, a Wi-Fi network, or plain text - and the extractor shows you whatever it finds, exactly as the code stores it, ready to copy. So if you need to get a mobile number out of a QR code, you'll see it come back as tel: followed by the number; a Wi-Fi code comes back with the network name and password in it. To go the other way and turn a number you already have into a code, use the phone number QR code generator.

Extractor or Scanner - Which Do You Need?

Both read your code with the same engine. Use this extractor when you have an image file of a QR code and want to pull the link out of it and regenerate a clean, print-ready copy. If you'd rather point a camera at a code or quickly read where one goes, the QR code scanner leads with your camera instead.

Get More Than a Copy - Make It Editable

The QR code you extract here is static: the link is baked into the pattern, so it can't be changed once it's printed. That's perfect when the destination is permanent. But if you want to update where the code points after it's printed - or see how many people scan it, and where - dynamic QR codes let you swap the destination any time and track every scan without reprinting a thing.