A café’s QR code is scratched, a label is stained, or a logo covers the centre. You point a camera at it and the link still appears. The scanner has not guessed the missing pixels by recognising the website. The code contains extra information designed to help reconstruct damaged data.
This feature is called error correction. It is one reason QR codes work outside ideal laboratory conditions, where printed labels encounter folds, dirt and imperfect cameras.
The square contains more than the message
A QR code encodes data into a pattern of small light and dark modules. It also contains structural features that help a reader locate and interpret the symbol. The three prominent corner patterns are part of that navigation system.
The encoded message is accompanied by redundant information calculated using error-correcting mathematics. QR codes use Reed–Solomon error correction, which allows a decoder to recover certain missing or incorrect codewords within its capacity.
A simple analogy is sending a message with carefully designed checking information. The receiver can use relationships among the surviving pieces to identify and reconstruct some errors. The real mathematics is more efficient than merely printing every word twice.
There are different protection levels
DENSO WAVE describes four error-correction levels commonly called L, M, Q and H. Higher levels provide greater recovery capacity but consume more of the symbol’s space, leaving less capacity for the message at a given size.
This creates a trade-off. A short link with substantial correction can be fairly robust. Packing more information into the same physical area can require smaller modules or different settings, making print quality and camera resolution increasingly important.
The often-quoted recovery percentages concern codewords under the coding scheme. They do not mean any arbitrary percentage of the visible square can safely be removed. The position and distribution of damage matter.
Why a small mark can still break it
A scanner first has to find and interpret the code. Damage to locating patterns, poor contrast or the loss of the blank margin can interfere before error correction gets a useful chance to work.
The surrounding clear area is called a quiet zone. DENSO WAVE specifies a four-module margin around a standard QR symbol. Crowding it with text or graphics can make the symbol harder to distinguish from its surroundings.
Blur, reflections and an oblique angle can also turn apparently intact modules into uncertain readings. A damaged code that works on one camera may fail on another because the input image is different.
The logo spends part of the allowance
Placing a logo over a QR code deliberately obscures some information. Error correction can sometimes tolerate it, but that uses capacity which might otherwise compensate for dirt or printing defects. A successful test once is not a guarantee under every condition.
A decoded code can also point to a website that no longer works. Error correction protects the encoded data, not the continued existence or trustworthiness of the destination.
The surprising resilience is therefore planned redundancy, not magic. The square carries enough extra structure to survive some damage, while still depending on readable patterns, appropriate sizing and a functioning destination.