Barcode Decoder

A barcode decoder is a tool that reads and decodes barcode images back into their original content. By uploading an image containing a barcode, the tool automatically recognizes and decodes the stored numeric or text information. It supports common 1D barcode formats including CODE128, EAN-13, EAN-8, CODE39, UPC-A, UPC-E, Codabar, and ITF, as well as 2D formats like QR Code, Data Matrix, PDF417, and Aztec. This tool runs entirely in your browser - your images never leave your device, ensuring complete privacy. Before publishing or submitting, open the output and check readability, cropping, resolution, ordering, and missing content.

A 1D barcode (UPC, EAN, Code 39, Code 128, ITF, Codabar) is a sequence of parallel bars and spaces whose widths encode digits, letters, or the full ASCII range. A 2D barcode (QR Code, Data Matrix, PDF417, Aztec) is a 2D pattern of modules (cells) arranged on a quiet zone, usually with three position-detection patterns in the corners for orientation. Decoding a barcode is the reverse: image → preprocess → locate the symbol → measure bar/space widths (or sample the 2D grid) → look up the encoding table → apply a checksum to validate the result. The image pipeline matters. The source image goes through grayscale conversion (luma = 0.299 R + 0.587 G + 0.114 B, the Rec. 601 weights), then a global or adaptive threshold (Otsu's method picks a threshold that maximises inter-class variance, which works for most well-lit scans; Sauvola's adaptive threshold handles uneven lighting better). Edge detection (Sobel kernel: 3×3 Gx and Gy) finds vertical bar transitions; a Hough transform on those lines gives orientation; connected-component analysis then finds the bounding rectangle of the symbol. The quiet zone (the mandatory blank margin around the symbol, 10× the X-dimension for QR Code, 9× the bar width for Code 128) is verified before the symbol is accepted as a valid candidate. For 1D codes, the decoder measures the X-dimension (the width of the narrowest bar, the unit length) and reads bar-space widths as multiples of X. A Code 39 character is 9 bars + 9 spaces (one of which is an inter-character gap); 3 of the 9 elements are wide (representing 1) and 6 are narrow (representing 0). A Code 128 character is 11 modules: a start pattern, 6 encoded bars/spaces, and a stop pattern. Each character is therefore 11 modules; 11 × character_count + 13 (for start + stop + check + quiet) gives the symbol width. Code 128 has three encoding sets: A (uppercase + control), B (uppercase + lowercase), C (paired digits 00-99) — set-switch codes let the encoder pick the most efficient set per substring. For 2D codes, the algorithm differs. QR Code uses Reed-Solomon error correction (EC level L/M/Q/H adds 7%, 15%, 25%, 30% redundancy) and a configurable masking pattern (one of 8; the encoder picks the one that minimises the penalty score). After masking, the bit stream is placed on a square grid of 21×21 (Version 1) up to 177×177 (Version 40) modules; the three 7×7 finder patterns in the corners and the 5×5 alignment patterns at the inner corners provide orientation and scale. Data Matrix uses a similar layout with an L-shaped solid border on two sides and alternating dark/light on the other two. PDF417 is a stacked linear code (17 modules wide, up to 90 rows) used on US driver's licences and airline boarding passes. This page uses the ZXing ('Zebra Crossing') library, an open-source Apache-2.0 multi-format decoder originally from the Google ZXing project. ZXing runs the full pipeline (locate → preprocess → decode) and tries each known symbology in turn. It supports UPC-A, UPC-E, EAN-8, EAN-13, Code 39, Code 93, Code 128, ITF, Codabar, QR Code, Data Matrix, PDF417, and Aztec. For 2D codes the success rate in good lighting is ~99% on a 1080p photo; for 1D codes it's lower (~85-95%) because perspective, blur, and glare distort the bar widths. This tool only accepts image input (file upload, drag-and-drop, or clipboard paste); it does not provide live camera scanning. Checksums matter. UPC/EAN use a single modulo-10 check digit (sum of odd-position digits × 3 + sum of even-position digits, mod 10, subtracted from 10). Code 128 uses a modulo-103 check character. Code 39 does not require a check digit but the mod-43 check is recommended for logistics. QR Code uses Reed-Solomon EC; if the EC decoder finds more errors than the code can correct, decode fails. The page surfaces decode failures with the reason (checksum mismatch, quiet zone too small, mask conflict) when the library exposes it.

• 1D barcodes (UPC, EAN, Code 39, Code 128, ITF) encode data in bar/space widths; 2D codes (QR Code, Data Matrix, PDF417, Aztec) encode in a 2D grid of modules. Both need a quiet zone of ~10× the unit width to be valid.
• Image pipeline: Rec. 601 grayscale (0.299R + 0.587G + 0.114B) → Otsu or Sauvola threshold → Sobel edge detection → Hough transform for orientation → connected-component analysis to find the bounding box of the symbol.
• Code 128 has 3 encoding sets: A (uppercase + control), B (uppercase + lowercase), C (paired digits 00-99). Set-switch codes let the encoder pick the most compact set per substring, packing UPC-A '123456789012' tighter than Code 39 would.
• Code 39: 9 bars + 9 spaces per character, 3 wide (1) and 6 narrow (0); start/stop character '*'. Optional mod-43 check digit. Widely used in automotive, defence, and US government (MIL-STD-1189).
• Code 93 is the predecessor of Code 128, with 9 modules per character (6 data + start/stop) and 47 unique characters; superseded by Code 128 in most applications but still seen in some Canadian pharmacy and USPS applications.
• QR Code: Reed-Solomon EC at L/M/Q/H levels (7%/15%/25%/30% redundancy), 8 masking patterns, 21×21 to 177×177 modules (Version 1 to 40). Three 7×7 finder patterns in the corners provide orientation; 5×5 alignment patterns on larger versions correct perspective distortion.
• Data Matrix: L-shaped solid border on two sides, alternating dark/light on the other two; used for small-item marking in electronics and pharmaceuticals (ISO/IEC 16022).
• ZXing ('Zebra Crossing', Apache-2.0, Google): the open-source multi-format decoder used here. Supports UPC-A/E, EAN-8/13, Code 39/93/128, ITF, Codabar, QR Code, Data Matrix, PDF417, and Aztec. This tool only decodes from image files (upload, drag-and-drop, or clipboard paste); it does not provide live camera scanning.
• Checksums: UPC/EAN use mod-10, Code 128 uses mod-103, Code 39 uses optional mod-43, QR Code uses Reed-Solomon EC. A checksum mismatch tells you the image is corrupted, not that the data is wrong.