libzbar: efficient cffi-based bindings for zbar

Python cffi-based bindings for zbar (http://zbar.sourceforge.net/) designed to be as efficient as possible.

Python 2, Python 3, and PyPy compatible!

Note: currently a work in progress. Entirely functional and production-ready (see NumPy note), but missing some configuration options.

Installation

$ pip install libzbar-cffi

Examples

>>> import libzbar as zb

# Images can be loaded from PIL images:
>>> from PIL import Image
>>> im = Image.open("test/qr-numeric.png")
>>> zb.Image.from_im(im).scan()
[<Symbol type=ZBAR_QRCODE quality=1 data='12345' locator=[(12, 12), (12, 75), (75, 75), (75, 12)]>]

# And from NumPy arrays. Note: if the array has an 8-bit data type (ex,
# uint8) a pointer to its data will be passed directly into zbar making
# this method very efficient.
>>> import numpy as np
>>> zb.Image.from_np(im.size, np.array(im)).scan()
[<Symbol type=ZBAR_QRCODE quality=1 data='12345' locator=[(12, 12), (12, 75), (75, 75), (75, 12)]>]

# By default, only QR codes will be scanned. Other symbol types can be
# scanned using the ``symbol_type`` argument (see ``libzbar.symbol_types``),
# or ``0`` for "all symbol types":
>>> ean13 = zb.Image.from_im(Image.open("test/ean13-example.png"))
>>> ean13.scan()
[]
>>> ean13.scan(symbol_type=0)
[<Symbol type=ZBAR_EAN13 quality=449 data='0012345678905' locator=[(30, 23), ..., (30, 247)]>]

NumPy Note

When using the Image.from_np(…) constructor a pointer to the underlying array is passed directly to zbar. The array's size and dtype are sanity checked, but at the moment the ctypes flags are ignored, so Strange Things might happen if the underyling array isn't a straight forward uint8_t[size[0] * size[1]].

Additionally, the array should not be changed for the lifecycle of the Image, as this could invalidate the data pointer.