@mbtech-nl/bitmap

Shared 1bpp bitmap rendering core for the MBTech thermal-label ecosystem. Powers @thermal-label/labelmanager-ts, @thermal-label/labelwriter-ts, @thermal-label/brother-ql-ts, @labelkit/designer-core. Pure TypeScript, zero runtime dependencies, no DOM APIs.

📖 Full documentation: mbtech-nl.github.io/bitmap — visual dither gallery, tone control demos, multi-plane walkthrough, interactive in-browser playground, full API reference.

Install

pnpm add @mbtech-nl/bitmap

Quick example

import{renderText,renderImage}from'@mbtech-nl/bitmap';consttextBitmap=renderText('HELLO',{scaleX: 2,scaleY: 2});constimageBitmap=renderImage(rawImageData,{dither: 'atkinson',rotate: 90});

For multi-colour printers (Brother QL-800 with red/black tape, two-colour DYMO/Zebra), see renderMultiPlaneImage.

Choosing a dither method

Content type	Recommended
Text, QR, barcodes, line drawings	dither: false (hard threshold)
Logos with sharp fills	dither: false, or 'atkinson' if anti-aliased
Photos — smoothest gradient	dither: 'jarvis-judice-ninke' or 'stucki'
Photos — crispest, most contrast	dither: 'atkinson'
Photos — balanced default	dither: 'floyd-steinberg'
Repeating patterns, posters, retro look	dither: 'bayer4' or 'bayer8'
Speed-critical inner loop	dither: 'bayer4' (no error propagation)

For side-by-side visuals of every method, see the dithering guide.

Also useful for…

Although built for thermal printers, the same primitives fit any pipeline that wants a clean 1bpp input:

• OCR / document AI — Tesseract and most classical OCR backends prefer binarised input. renderImage with autoLevels: true handles low-contrast scans; dither: false (hard threshold) preserves character edges.
• Barcode / QR preprocessing — same story; binarisation + nearest-neighbour scaling is exactly what decoders want.
• Handwriting recognition / layout analysis — feed normalised binary frames into downstream classical or ML models.
• Multi-colour mask separation — renderMultiPlaneImage produces mutually-exclusive masks per palette colour, useful as a deterministic colour-quantisation step before downstream processing.

Not a fit for general computer-vision preprocessing (CNNs, ViTs, CLIP) — those want full tonal range with bilinear/bicubic resampling, the opposite of what this lib produces. Reach for torchvision/albumentations/PIL there.

API at a glance

• renderText, measureText — render ASCII into 1bpp using the bundled 8×8 font.
• renderImage — RGBA → 1bpp with threshold or one of six dither methods, plus autoLevels, gamma, luminanceWeights.
• renderMultiPlaneImage — RGBA → one 1bpp plane per palette entry, mutually exclusive (multi-colour printers).
• Transforms: rotateBitmap, flipHorizontal, flipVertical, invertBitmap, scaleBitmap, cropBitmap, stackBitmaps, padBitmap.
• Encoding helpers: getRow, iterRows, bytesPerRow, createBitmap, bitmapEquals, getPixel.

Full signatures with examples: API reference.

Bit layout

LabelBitmap stores pixels row-major and MSB-first:

• Byte index: y * ceil(widthPx / 8) + floor(x / 8)
• Bit index in byte: 7 - (x % 8)
• Bit value: 1 = black, 0 = white

Trailing bits in the last byte of each row are always zeroed.

Browser support

Pure TypeScript, zero runtime dependencies, no DOM APIs. Runs in modern browsers and Node.js where Uint8Array is available — including the interactive playground, which uses the same code that ships to npm.

Font

Bundled default8x8 uses IBM-derived CP437-compatible 8×8 glyphs for ASCII 0x20..0x7F.

• Source: dhepper/font8x8
• License: Public Domain

Contributing

mbtech-nl/.github/CONTRIBUTING.md

License

MIT