glymur/glymur/jp2k.py
John Evans 4686c40c57 fix test failure, not a real bug, closes #316
Had assumed that the error was due to parse_options not being properly
reset, but that was not the case.  Seems to have just been bad expected
data.
2015-01-08 12:07:10 -05:00

1924 lines
69 KiB
Python

"""This file is part of glymur, a Python interface for accessing JPEG 2000.
http://glymur.readthedocs.org
Copyright 2013 John Evans
License: MIT
"""
import sys
# Exitstack not found in contextlib in 2.7
if sys.hexversion >= 0x03030000:
from contextlib import ExitStack
from itertools import filterfalse
else:
from contextlib2 import ExitStack
from itertools import ifilterfalse as filterfalse
from collections import Counter
import ctypes
import math
import os
import re
import struct
from uuid import UUID
import warnings
import numpy as np
from .codestream import Codestream
from . import core, version
from .jp2box import (Jp2kBox, JPEG2000SignatureBox, FileTypeBox,
JP2HeaderBox, ColourSpecificationBox,
ContiguousCodestreamBox, ImageHeaderBox)
from .lib import openjpeg as opj, openjp2 as opj2, c as libc
class Jp2k(Jp2kBox):
"""JPEG 2000 file.
Attributes
----------
filename : str
The path to the JPEG 2000 file.
box : sequence
List of top-level boxes in the file. Each box may in turn contain
its own list of boxes. Will be empty if the file consists only of a
raw codestream.
shape : tuple
Size of the image.
Properties
----------
ignore_pclr_cmap_cdef : bool
whether or not to ignore the pclr, cmap, or cdef boxes during any
color transformation, defaults to False.
layer : int
zero-based number of quality layer to decode
verbose : bool
whether or not to print informational messages produced by the
OpenJPEG library, defaults to false
codestream : object
JP2 or J2K codestream object
Examples
--------
>>> import glymur
>>> jfile = glymur.data.nemo()
>>> jp2 = glymur.Jp2k(jfile)
>>> jp2.shape
(1456, 2592, 3)
>>> image = jp2[:]
>>> image.shape
(1456, 2592, 3)
Read a lower resolution thumbnail.
>>> thumbnail = jp2[::2, ::2]
>>> thumbnail.shape
(728, 1296, 3)
"""
def __init__(self, filename, data=None, shape=None, **kwargs):
"""
Only the filename parameter is required in order to read a JPEG 2000
file.
Parameters
----------
filename : str or file
the path to JPEG 2000 file
image_data : ndarray, optional
image data to be written
shape : tuple
size of image data, only required when image_data is not provided
cbsize : tuple, optional
code block size (DY, DX)
cinema2k : int, optional
frames per second, either 24 or 48
cinema4k : bool, optional
set to True to specify Cinema4K mode, defaults to false
colorspace : str, optional
either 'rgb' or 'gray'
cratios : iterable
compression ratios for successive layers
eph : bool, optional
if true, write SOP marker after each header packet
grid_offset : tuple, optional
offset (DY, DX) of the origin of the image in the reference grid
irreversible : bool, optional
if true, use the irreversible DWT 9-7 transform
mct : bool, optional
specifies usage of the multi component transform, if not
specified, defaults to True if the colorspace is RGB
modesw : int, optional
mode switch
1 = BYPASS(LAZY)
2 = RESET
4 = RESTART(TERMALL)
8 = VSC
16 = ERTERM(SEGTERM)
32 = SEGMARK(SEGSYM)
numres : int, optional
number of resolutions
prog : str, optional
progression order, one of "LRCP" "RLCP", "RPCL", "PCRL", "CPRL"
psnr : iterable, optional
different PSNR for successive layers
psizes : list, optional
list of precinct sizes, each precinct size tuple is defined in
(height x width)
sop : bool, optional
if true, write SOP marker before each packet
subsam : tuple, optional
subsampling factors (dy, dx)
tilesize : tuple, optional
numeric tuple specifying tile size in terms of (numrows, numcols),
not (X, Y)
verbose : bool, optional
print informational messages produced by the OpenJPEG library
"""
Jp2kBox.__init__(self)
self.filename = filename
self.box = []
self._codec_format = None
self._colorspace = None
self._layer = 0
self._codestream = None
if data is not None:
self._shape = data.shape
else:
self._shape = shape
self._ignore_pclr_cmap_cdef = False
self._verbose = False
# Parse the file for JP2/JPX contents only if we are reading it.
if data is None and shape is None:
self.parse()
elif data is not None:
self._write(data, **kwargs)
@property
def ignore_pclr_cmap_cdef(self):
return self._ignore_pclr_cmap_cdef
@ignore_pclr_cmap_cdef.setter
def ignore_pclr_cmap_cdef(self, ignore_pclr_cmap_cdef):
self._ignore_pclr_cmap_cdef = ignore_pclr_cmap_cdef
@property
def layer(self):
return self._layer
@layer.setter
def layer(self, layer):
if version.openjpeg_version_tuple[0] < 2:
msg = "Layer property not supported unless the version of "
msg += "OpenJPEG is 2.0 or higher."
raise RuntimeError(msg)
self._layer = layer
@property
def codestream(self):
if self._codestream is None:
self._codestream = self.get_codestream(header_only=True)
return self._codestream
@codestream.setter
def codestream(self, the_codestream):
self._codestream = the_codestream
@property
def verbose(self):
return self._verbose
@verbose.setter
def verbose(self, verbose):
self._verbose = verbose
@property
def shape(self):
if self._shape is not None:
return self._shape
if self._codec_format == opj2.CODEC_J2K:
# get the image size from the codestream
cstr = self.codestream
height = cstr.segment[1].ysiz
width = cstr.segment[1].xsiz
num_components = len(cstr.segment[1].xrsiz)
else:
# try to get the image size from the IHDR box
jp2h = [box for box in self.box if box.box_id == 'jp2h'][0]
ihdr = [box for box in jp2h.box if box.box_id == 'ihdr'][0]
height, width = ihdr.height, ihdr.width
num_components = ihdr.num_components
if num_components == 1:
# but if there is a PCLR box, then we need to check that as
# well, as that turns a single-channel image into a
# multi-channel image
pclr = [box for box in jp2h.box if box.box_id == 'pclr']
if len(pclr) > 0:
num_components = len(pclr[0].signed)
if num_components == 1:
self.shape = (height, width)
else:
self.shape = (height, width, num_components)
return self._shape
@shape.setter
def shape(self, shape):
self._shape = shape
def __repr__(self):
msg = "glymur.Jp2k('{0}')".format(self.filename)
return msg
def __str__(self):
metadata = ['File: ' + os.path.basename(self.filename)]
if len(self.box) > 0:
for box in self.box:
metadata.append(str(box))
else:
metadata.append(str(self.codestream))
return '\n'.join(metadata)
def parse(self):
"""Parses the JPEG 2000 file.
Raises
------
IOError
The file was not JPEG 2000.
"""
self.length = os.path.getsize(self.filename)
with open(self.filename, 'rb') as fptr:
# Make sure we have a JPEG2000 file. It could be either JP2 or
# J2C. Check for J2C first, single box in that case.
read_buffer = fptr.read(2)
signature, = struct.unpack('>H', read_buffer)
if signature == 0xff4f:
self._codec_format = opj2.CODEC_J2K
# That's it, we're done. The codestream object is only
# produced upon explicit request.
return
self._codec_format = opj2.CODEC_JP2
# Should be JP2.
# First 4 bytes should be 12, the length of the 'jP ' box.
# 2nd 4 bytes should be the box ID ('jP ').
# 3rd 4 bytes should be the box signature (13, 10, 135, 10).
fptr.seek(0)
read_buffer = fptr.read(12)
values = struct.unpack('>I4s4B', read_buffer)
box_length = values[0]
box_id = values[1]
signature = values[2:]
if (((box_length != 12) or (box_id != b'jP ') or
(signature != (13, 10, 135, 10)))):
msg = '{0} is not a JPEG 2000 file.'.format(self.filename)
raise IOError(msg)
# Back up and start again, we know we have a superbox (box of
# boxes) here.
fptr.seek(0)
self.box = self.parse_superbox(fptr)
self._validate()
def _validate(self):
"""Validate the JPEG 2000 outermost superbox.
"""
# A jp2-branded file cannot contain an "any ICC profile
ftyp = self.box[1]
if ftyp.brand == 'jp2 ':
jp2h = [box for box in self.box if box.box_id == 'jp2h'][0]
colrs = [box for box in jp2h.box if box.box_id == 'colr']
for colr in colrs:
if colr.method not in (core.ENUMERATED_COLORSPACE,
core.RESTRICTED_ICC_PROFILE):
msg = "Color Specification box method must specify either "
msg += "an enumerated colorspace or a restricted ICC "
msg += "profile if the file type box brand is 'jp2 '."
warnings.warn(msg)
def _set_cinema_params(self, cinema_mode, fps):
"""Populate compression parameters structure for cinema2K.
Parameters
----------
params : ctypes struct
Corresponds to compression parameters structure used by the
library.
cinema_mode : str
Either 'cinema2k' or 'cinema4k'
fps : int
Frames per second, should be either 24 or 48.
"""
if re.match("1.5|2.0.0", version.openjpeg_version) is not None:
msg = "Writing Cinema2K or Cinema4K files is not supported with "
msg += 'openjpeg library versions less than 2.0.1.'
raise IOError(msg)
# Cinema modes imply MCT.
self._cparams.tcp_mct = 1
if cinema_mode == 'cinema2k':
if fps not in [24, 48]:
raise IOError('Cinema2K frame rate must be either 24 or 48.')
if re.match("2.0", version.openjpeg_version) is not None:
# 2.0 API
if fps == 24:
self._cparams.cp_cinema = core.OPJ_CINEMA2K_24
else:
self._cparams.cp_cinema = core.OPJ_CINEMA2K_48
else:
# 2.1 API
if fps == 24:
self._cparams.rsiz = core.OPJ_PROFILE_CINEMA_2K
self._cparams.max_comp_size = core.OPJ_CINEMA_24_COMP
self._cparams.max_cs_size = core.OPJ_CINEMA_24_CS
else:
self._cparams.rsiz = core.OPJ_PROFILE_CINEMA_2K
self._cparams.max_comp_size = core.OPJ_CINEMA_48_COMP
self._cparams.max_cs_size = core.OPJ_CINEMA_48_CS
else:
# cinema4k
if re.match("2.0", version.openjpeg_version) is not None:
# 2.0 API
self._cparams.cp_cinema = core.OPJ_CINEMA4K_24
else:
# 2.1 API
self._cparams.rsiz = core.OPJ_PROFILE_CINEMA_4K
def _populate_cparams(self, img_array, **kwargs):
"""Directs processing of write method arguments.
Parameters
----------
img_array : ndarray
image data to be written to file
kwargs : dictionary
non-image keyword inputs provided to write method
"""
if ((('cinema2k' in kwargs or 'cinema4k' in kwargs) and
(len(set(kwargs)) > 1))):
msg = "Cannot specify cinema2k/cinema4k along with other options."
raise IOError(msg)
if 'cratios' in kwargs and 'psnr' in kwargs:
msg = "Cannot specify cratios and psnr together."
raise IOError(msg)
if version.openjpeg_version_tuple[0] == 1:
cparams = opj.set_default_encoder_parameters()
else:
cparams = opj2.set_default_encoder_parameters()
outfile = self.filename.encode()
num_pad_bytes = opj2.PATH_LEN - len(outfile)
outfile += b'0' * num_pad_bytes
cparams.outfile = outfile
if self.filename[-4:].endswith(('.jp2', '.JP2')):
cparams.codec_fmt = opj2.CODEC_JP2
else:
cparams.codec_fmt = opj2.CODEC_J2K
# Set defaults to lossless to begin.
cparams.tcp_rates[0] = 0
cparams.tcp_numlayers = 1
cparams.cp_disto_alloc = 1
if 'irreversible' in kwargs and kwargs['irreversible'] is True:
cparams.irreversible = 1
if 'cinema2k' in kwargs:
self._cparams = cparams
self._set_cinema_params('cinema2k', kwargs['cinema2k'])
return
if 'cinema4k' in kwargs:
self._cparams = cparams
self._set_cinema_params('cinema4k', kwargs['cinema4k'])
return
if 'cbsize' in kwargs:
cparams.cblockw_init = kwargs['cbsize'][1]
cparams.cblockh_init = kwargs['cbsize'][0]
if 'cratios' in kwargs:
cparams.tcp_numlayers = len(kwargs['cratios'])
for j, cratio in enumerate(kwargs['cratios']):
cparams.tcp_rates[j] = cratio
cparams.cp_disto_alloc = 1
if 'eph' in kwargs:
cparams.csty |= 0x04
if 'grid_offset' in kwargs:
cparams.image_offset_x0 = kwargs['grid_offset'][1]
cparams.image_offset_y0 = kwargs['grid_offset'][0]
if 'modesw' in kwargs:
for shift in range(6):
power_of_two = 1 << shift
if kwargs['modesw'] & power_of_two:
cparams.mode |= power_of_two
if 'numres' in kwargs:
cparams.numresolution = kwargs['numres']
if 'prog' in kwargs:
prog = kwargs['prog'].upper()
cparams.prog_order = core.PROGRESSION_ORDER[prog]
if 'psnr' in kwargs:
cparams.tcp_numlayers = len(kwargs['psnr'])
for j, snr_layer in enumerate(kwargs['psnr']):
cparams.tcp_distoratio[j] = snr_layer
cparams.cp_fixed_quality = 1
if 'psizes' in kwargs:
for j, (prch, prcw) in enumerate(kwargs['psizes']):
cparams.prcw_init[j] = prcw
cparams.prch_init[j] = prch
cparams.csty |= 0x01
cparams.res_spec = len(kwargs['psizes'])
if 'sop' in kwargs:
cparams.csty |= 0x02
if 'subsam' in kwargs:
cparams.subsampling_dy = kwargs['subsam'][0]
cparams.subsampling_dx = kwargs['subsam'][1]
if 'tilesize' in kwargs:
cparams.cp_tdx = kwargs['tilesize'][1]
cparams.cp_tdy = kwargs['tilesize'][0]
cparams.tile_size_on = opj2.TRUE
try:
mct = kwargs['mct']
if mct and self._colorspace == opj2.CLRSPC_GRAY:
msg = "Cannot specify usage of the multi component transform "
msg += "if the colorspace is gray."
raise IOError(msg)
cparams.tcp_mct = 1 if mct else 0
except KeyError:
# If the multi component transform was not specified, we infer
# that it should be used if the color space is RGB.
if self._colorspace == opj2.CLRSPC_SRGB:
cparams.tcp_mct = 1
else:
cparams.tcp_mct = 0
self._validate_compression_params(img_array, cparams, **kwargs)
self._cparams = cparams
def _write(self, img_array, verbose=False, **kwargs):
"""Write image data to a JP2/JPX/J2k file. Intended usage of the
various parameters follows that of OpenJPEG's opj_compress utility.
This method can only be used to create JPEG 2000 images that can fit
in memory.
"""
if re.match("0|1.[0-4]", version.openjpeg_version) is not None:
msg = "You must have at least version 1.5 of OpenJPEG "
msg += "in order to write images."
raise RuntimeError(msg)
self._determine_colorspace(**kwargs)
self._populate_cparams(img_array, **kwargs)
if opj2.OPENJP2 is not None:
self._write_openjp2(img_array, verbose=verbose)
else:
self._write_openjpeg(img_array, verbose=verbose)
def _write_openjpeg(self, img_array, verbose=False):
"""
Write JPEG 2000 file using OpenJPEG 1.5 interface.
"""
if img_array.ndim == 2:
# Force the image to be 3D. Just makes things easier later on.
img_array = img_array.reshape(img_array.shape[0],
img_array.shape[1],
1)
self._populate_comptparms(img_array)
with ExitStack() as stack:
image = opj.image_create(self._comptparms, self._colorspace)
stack.callback(opj.image_destroy, image)
numrows, numcols, numlayers = img_array.shape
# set image offset and reference grid
image.contents.x0 = self._cparams.image_offset_x0
image.contents.y0 = self._cparams.image_offset_y0
image.contents.x1 = (image.contents.x0
+ (numcols - 1) * self._cparams.subsampling_dx
+ 1)
image.contents.y1 = (image.contents.y0
+ (numrows - 1) * self._cparams.subsampling_dy
+ 1)
# Stage the image data to the openjpeg data structure.
for k in range(0, numlayers):
layer = np.ascontiguousarray(img_array[:, :, k],
dtype=np.int32)
dest = image.contents.comps[k].data
src = layer.ctypes.data
ctypes.memmove(dest, src, layer.nbytes)
cinfo = opj.create_compress(self._cparams.codec_fmt)
stack.callback(opj.destroy_compress, cinfo)
# Setup the info, warning, and error handlers.
# Always use the warning and error handler. Use of an info
# handler is optional.
event_mgr = opj.EventMgrType()
_info_handler = _INFO_CALLBACK if verbose else None
event_mgr.info_handler = _info_handler
event_mgr.warning_handler = ctypes.cast(_WARNING_CALLBACK,
ctypes.c_void_p)
event_mgr.error_handler = ctypes.cast(_ERROR_CALLBACK,
ctypes.c_void_p)
opj.setup_encoder(cinfo, ctypes.byref(self._cparams), image)
cio = opj.cio_open(cinfo)
stack.callback(opj.cio_close, cio)
if not opj.encode(cinfo, cio, image):
raise IOError("Encode error.")
pos = opj.cio_tell(cio)
blob = ctypes.string_at(cio.contents.buffer, pos)
fptr = open(self.filename, 'wb')
stack.callback(fptr.close)
fptr.write(blob)
self.parse()
def _validate_compression_params(self, img_array, cparams, **kwargs):
"""Check that the compression parameters are valid.
Parameters
----------
img_array : ndarray
Image data to be written to file.
cparams : CompressionParametersType(ctypes.Structure)
Corresponds to cparameters_t type in openjp2 headers.
"""
# Cannot specify a colorspace with J2K.
if cparams.codec_fmt == opj2.CODEC_J2K and 'colorspace' in kwargs:
msg = 'Do not specify a colorspace when writing a raw '
msg += 'codestream.'
raise IOError(msg)
# Code block size
code_block_specified = False
if cparams.cblockw_init != 0 and cparams.cblockh_init != 0:
# These fields ARE zero if uninitialized.
width = cparams.cblockw_init
height = cparams.cblockh_init
code_block_specified = True
if height * width > 4096 or height < 4 or width < 4:
msg = "Code block area cannot exceed 4096. "
msg += "Code block height and width must be larger than 4."
raise IOError(msg)
if ((math.log(height, 2) != math.floor(math.log(height, 2)) or
math.log(width, 2) != math.floor(math.log(width, 2)))):
msg = "Bad code block size ({0}, {1}), "
msg += "must be powers of 2."
raise IOError(msg.format(height, width))
# Precinct size
if cparams.res_spec != 0:
# precinct size was not specified if this field is zero.
for j in range(cparams.res_spec):
prch = cparams.prch_init[j]
prcw = cparams.prcw_init[j]
if j == 0 and code_block_specified:
height, width = cparams.cblockh_init, cparams.cblockw_init
if height * 2 > prch or width * 2 > prcw:
msg = "Highest Resolution precinct size must be at "
msg += "least twice that of the code block dimensions."
raise IOError(msg)
if ((math.log(prch, 2) != math.floor(math.log(prch, 2)) or
math.log(prcw, 2) != math.floor(math.log(prcw, 2)))):
msg = "Bad precinct sizes ({0}, {1}), "
msg += "must be powers of 2."
raise IOError(msg.format(prch, prcw))
# What would the point of 1D images be?
if img_array.ndim == 1 or img_array.ndim > 3:
msg = "{0}D imagery is not allowed.".format(img_array.ndim)
raise IOError(msg)
if re.match("2.0.0", version.openjpeg_version) is not None:
if (((img_array.ndim != 2) and
(img_array.shape[2] != 1 and img_array.shape[2] != 3))):
msg = "Writing images is restricted to single-channel "
msg += "greyscale images or three-channel RGB images when "
msg += "the OpenJPEG library version is the official 2.0.0 "
msg += "release."
raise IOError(msg)
if img_array.dtype != np.uint8 and img_array.dtype != np.uint16:
msg = "Only uint8 and uint16 datatypes are currently supported "
msg += "when writing."
raise RuntimeError(msg)
def _determine_colorspace(self, colorspace=None, **kwargs):
"""Determine the colorspace from the supplied inputs.
Parameters
----------
colorspace : str, optional
Either 'rgb' or 'gray'.
"""
if colorspace is None:
# Must infer the colorspace from the image dimensions.
if len(self.shape) < 3:
# A single channel image is grayscale.
self._colorspace = opj2.CLRSPC_GRAY
elif self.shape[2] == 1 or self.shape[2] == 2:
# A single channel image or an image with two channels is going
# to be greyscale.
self._colorspace = opj2.CLRSPC_GRAY
else:
# Anything else must be RGB, right?
self._colorspace = opj2.CLRSPC_SRGB
else:
if colorspace.lower() not in ('rgb', 'grey', 'gray'):
msg = 'Invalid colorspace "{0}"'.format(colorspace)
raise IOError(msg)
elif colorspace.lower() == 'rgb' and self.shape[2] < 3:
msg = 'RGB colorspace requires at least 3 components.'
raise IOError(msg)
# Turn the colorspace from a string to the enumerated value that
# the library expects.
COLORSPACE_MAP = {'rgb': opj2.CLRSPC_SRGB,
'gray': opj2.CLRSPC_GRAY,
'grey': opj2.CLRSPC_GRAY,
'ycc': opj2.CLRSPC_YCC}
self._colorspace = COLORSPACE_MAP[colorspace.lower()]
def _write_openjp2(self, img_array, verbose=False):
"""
Write JPEG 2000 file using OpenJPEG 2.x interface.
"""
if img_array.ndim == 2:
# Force the image to be 3D. Just makes things easier later on.
numrows, numcols = img_array.shape
img_array = img_array.reshape(numrows, numcols, 1)
self._populate_comptparms(img_array)
with ExitStack() as stack:
image = opj2.image_create(self._comptparms, self._colorspace)
stack.callback(opj2.image_destroy, image)
self._populate_image_struct(image, img_array)
codec = opj2.create_compress(self._cparams.codec_fmt)
stack.callback(opj2.destroy_codec, codec)
if self._verbose or verbose:
info_handler = _INFO_CALLBACK
else:
info_handler = None
opj2.set_info_handler(codec, info_handler)
opj2.set_warning_handler(codec, _WARNING_CALLBACK)
opj2.set_error_handler(codec, _ERROR_CALLBACK)
opj2.setup_encoder(codec, self._cparams, image)
if re.match("2.0", version.openjpeg_version) is not None:
fptr = libc.fopen(self.filename, 'wb')
strm = opj2.stream_create_default_file_stream(fptr, False)
stack.callback(opj2.stream_destroy, strm)
stack.callback(libc.fclose, fptr)
else:
# Introduced in 2.1 devel series.
strm = opj2.stream_create_default_file_stream(self.filename,
False)
stack.callback(opj2.stream_destroy, strm)
opj2.start_compress(codec, image, strm)
opj2.encode(codec, strm)
opj2.end_compress(codec, strm)
# Refresh the metadata.
self.parse()
def append(self, box):
"""Append a JP2 box to the file in-place.
Parameters
----------
box : Jp2Box
Instance of a JP2 box. Only UUID and XML boxes can currently be
appended.
"""
if self._codec_format == opj2.CODEC_J2K:
msg = "Only JP2 files can currently have boxes appended to them."
raise IOError(msg)
if not ((box.box_id == 'xml ') or
(box.box_id == 'uuid' and
box.uuid == UUID('be7acfcb-97a9-42e8-9c71-999491e3afac'))):
msg = "Only XML boxes and XMP UUID boxes can currently be "
msg += "appended."
raise IOError(msg)
# Check the last box. If the length field is zero, then rewrite
# the length field to reflect the true length of the box.
with open(self.filename, 'rb') as ifile:
offset = self.box[-1].offset
ifile.seek(offset)
read_buffer = ifile.read(4)
box_length, = struct.unpack('>I', read_buffer)
if box_length == 0:
# Reopen the file in write mode and rewrite the length field.
true_box_length = os.path.getsize(ifile.name) - offset
with open(self.filename, 'r+b') as ofile:
ofile.seek(offset)
write_buffer = struct.pack('>I', true_box_length)
ofile.write(write_buffer)
# Can now safely append the box.
with open(self.filename, 'ab') as ofile:
box.write(ofile)
self.parse()
def wrap(self, filename, boxes=None):
"""Create a new JP2/JPX file wrapped in a new set of JP2 boxes.
This method is primarily aimed at wrapping a raw codestream in a set of
of JP2 boxes (turning it into a JP2 file instead of just a raw
codestream), or rewrapping a codestream in a JP2 file in a new "jacket"
of JP2 boxes.
Parameters
----------
filename : str
JP2 file to be created from a raw codestream.
boxes : list
JP2 box definitions to define the JP2 file format. If not
provided, a default ""jacket" is assumed, consisting of JP2
signature, file type, JP2 header, and contiguous codestream boxes.
A JPX file rewrapped without the boxes argument results in a JP2
file encompassing the first codestream.
Returns
-------
jp2 : Jp2k object
Newly wrapped Jp2k object.
Examples
--------
>>> import glymur, tempfile
>>> jfile = glymur.data.goodstuff()
>>> j2k = glymur.Jp2k(jfile)
>>> tfile = tempfile.NamedTemporaryFile(suffix='jp2')
>>> jp2 = j2k.wrap(tfile.name)
"""
if boxes is None:
boxes = self._get_default_jp2_boxes()
_validate_jp2_box_sequence(boxes)
with open(filename, 'wb') as ofile:
for box in boxes:
if box.box_id != 'jp2c':
box.write(ofile)
else:
self._write_wrapped_codestream(ofile, box)
ofile.flush()
jp2 = Jp2k(filename)
return jp2
def _write_wrapped_codestream(self, ofile, box):
"""Write wrapped codestream."""
# Codestreams require a bit more care.
# Am I a raw codestream?
if len(self.box) == 0:
# Yes, just write the codestream box header plus all
# of myself out to file.
ofile.write(struct.pack('>I', self.length + 8))
ofile.write(b'jp2c')
with open(self.filename, 'rb') as ifile:
ofile.write(ifile.read())
return
# OK, I'm a jp2/jpx file. Need to find out where the raw codestream
# actually starts.
offset = box.offset
if offset == -1:
if self.box[1].brand == 'jpx ':
msg = "The codestream box must have its offset and "
msg += "length attributes fully specified if the file "
msg += "type brand is JPX."
raise IOError(msg)
# Find the first codestream in the file.
jp2c = [_box for _box in self.box if _box.box_id == 'jp2c']
offset = jp2c[0].offset
# Ready to write the codestream.
with open(self.filename, 'rb') as ifile:
ifile.seek(offset)
# Verify that the specified codestream is right.
read_buffer = ifile.read(8)
L, T = struct.unpack_from('>I4s', read_buffer, 0)
if T != b'jp2c':
msg = "Unable to locate the specified codestream."
raise IOError(msg)
if L == 0:
# The length of the box is presumed to last until the end of
# the file. Compute the effective length of the box.
L = os.path.getsize(ifile.name) - ifile.tell() + 8
elif L == 1:
# The length of the box is in the XL field, a 64-bit value.
read_buffer = ifile.read(8)
L, = struct.unpack('>Q', read_buffer)
ifile.seek(offset)
read_buffer = ifile.read(L)
ofile.write(read_buffer)
def _get_default_jp2_boxes(self):
"""Create a default set of JP2 boxes."""
# Try to create a reasonable default.
boxes = [JPEG2000SignatureBox(),
FileTypeBox(),
JP2HeaderBox(),
ContiguousCodestreamBox()]
height = self.codestream.segment[1].ysiz
width = self.codestream.segment[1].xsiz
num_components = len(self.codestream.segment[1].xrsiz)
if num_components < 3:
colorspace = core.GREYSCALE
else:
if len(self.box) == 0:
# Best guess is SRGB
colorspace = core.SRGB
else:
# Take whatever the first jp2 header / color specification
# says.
jp2hs = [box for box in self.box if box.box_id == 'jp2h']
colorspace = jp2hs[0].box[1].colorspace
boxes[2].box = [ImageHeaderBox(height=height, width=width,
num_components=num_components),
ColourSpecificationBox(colorspace=colorspace)]
return boxes
def __setitem__(self, index, data):
"""
Slicing protocol.
"""
if ((isinstance(index, slice) and
(index.start is None and
index.stop is None and
index.step is None)) or (index is Ellipsis)):
# Case of jp2[:] = data, i.e. write the entire image.
#
# Should have a slice object where start = stop = step = None
self._write(data)
else:
msg = "Partial write operations are currently not allowed."
raise TypeError(msg)
def __getitem__(self, pargs):
"""
Slicing protocol.
"""
if len(self.shape) == 2:
numrows, numcols = self.shape
numbands = 1
else:
numrows, numcols, numbands = self.shape
if isinstance(pargs, int):
# Not a very good use of this protocol, but technically legal.
# This retrieves a single row.
row = pargs
area = (row, 0, row + 1, numcols)
return self._read(area=area).squeeze()
if pargs is Ellipsis:
# Case of jp2[...]
return self._read()
if isinstance(pargs, slice):
if (((pargs.start is None) and
(pargs.stop is None) and
(pargs.step is None))):
# Case of jp2[:]
return self._read()
# Corner case of jp2[x] where x is a slice object with non-null
# members. Just augment it with an ellipsis and let the code
# below handle it.
pargs = (pargs, Ellipsis)
if isinstance(pargs, tuple) and any(x is Ellipsis for x in pargs):
# Remove the first ellipsis we find.
rows = slice(0, numrows)
cols = slice(0, numcols)
bands = slice(0, numbands)
if pargs[0] is Ellipsis:
if len(pargs) == 2:
newindex = (rows, cols, pargs[1])
else:
newindex = (rows, pargs[1], pargs[2])
elif pargs[1] is Ellipsis:
if len(pargs) == 2:
newindex = (pargs[0], cols, bands)
else:
newindex = (pargs[0], cols, pargs[2])
else:
# Assume that we don't have 4D imagery, of course.
newindex = (pargs[0], pargs[1], bands)
# Run once again because it is possible that there's another
# Ellipsis object in the 2nd or 3rd position.
return self.__getitem__(newindex)
if isinstance(pargs, tuple) and any(isinstance(x, int) for x in pargs):
# Replace the first such integer argument, replace it with a slice.
lst = list(pargs)
predicate = lambda x: not isinstance(x[1], int)
g = filterfalse(predicate, enumerate(pargs))
idx = next(g)[0]
lst[idx] = slice(pargs[idx], pargs[idx] + 1)
newindex = tuple(lst)
# Invoke array-based slicing again, as there may be additional
# integer argument remaining.
data = self.__getitem__(newindex)
# Reduce dimensionality in the scalar dimension.
return np.squeeze(data, axis=idx)
# Assuming pargs is a tuple of slices from now on.
rows = pargs[0]
cols = pargs[1]
if len(pargs) == 2:
bands = slice(None, None, None)
else:
bands = pargs[2]
rows_step = 1 if rows.step is None else rows.step
cols_step = 1 if cols.step is None else cols.step
if rows_step != cols_step:
msg = "Row and column strides must be the same."
raise IndexError(msg)
# Ok, reduce layer step is the same in both xy directions, so just take
# one of them.
step = rows_step
# Check if the step size is a power of 2.
if np.abs(np.log2(step) - np.round(np.log2(step))) > 1e-6:
msg = "Row and column strides must be powers of 2."
raise IndexError(msg)
rlevel = np.int(np.round(np.log2(step)))
area = (0 if rows.start is None else rows.start,
0 if cols.start is None else cols.start,
numrows if rows.stop is None else rows.stop,
numcols if cols.stop is None else cols.stop
)
data = self._read(area=area, rlevel=rlevel)
if len(pargs) == 2:
return data
# Ok, 3 arguments in pargs.
return data[:, :, bands]
def _read(self, **kwargs):
"""Read a JPEG 2000 image.
Returns
-------
img_array : ndarray
The image data.
Raises
------
IOError
If the image has differing subsample factors.
"""
if version.openjpeg_version_tuple[0] < 2:
img = self._read_openjpeg(**kwargs)
else:
img = self._read_openjp2(**kwargs)
return img
def read(self, **kwargs):
"""
"""
# Read a JPEG 2000 image.
#
# Parameters
# ----------
# rlevel : int, optional
# Factor by which to rlevel output resolution. Use -1 to get the
# lowest resolution thumbnail. This is the only keyword option
# available to use when the OpenJPEG version is 1.5 or earlier.
# layer : int, optional
# Number of quality layer to decode.
# area : tuple, optional
# Specifies decoding image area,
# (first_row, first_col, last_row, last_col)
# tile : int, optional
# Number of tile to decode.
# verbose : bool, optional
# Print informational messages produced by the OpenJPEG library.
#
# Returns
# -------
# img_array : ndarray
# The image data.
#
# Raises
# ------
# IOError
# If the image has differing subsample factors.
if 'ignore_pclr_cmap_cdef' in kwargs:
self.ignore_pclr_cmap_cdef = kwargs['ignore_pclr_cmap_cdef']
warnings.warn("Use array-style slicing instead.", DeprecationWarning)
if version.openjpeg_version_tuple[0] < 2:
img = self._read_openjpeg(**kwargs)
else:
img = self._read_openjp2(**kwargs)
return img
def _subsampling_sanity_check(self):
"""Check for differing subsample factors.
"""
dxs = np.array(self.codestream.segment[1].xrsiz)
dys = np.array(self.codestream.segment[1].yrsiz)
if np.any(dxs - dxs[0]) or np.any(dys - dys[0]):
msg = "Components must all have the same subsampling factors "
msg += "to use this method. Please consider using OPENJP2 and "
msg += "the read_bands method instead."
raise RuntimeError(msg)
def _read_openjpeg(self, rlevel=0, verbose=False, area=None):
"""Read a JPEG 2000 image using libopenjpeg.
Parameters
----------
rlevel : int, optional
Factor by which to rlevel output resolution. Use -1 to get the
lowest resolution thumbnail.
verbose : bool, optional
Print informational messages produced by the OpenJPEG library.
area : tuple, optional
Specifies decoding image area,
(first_row, first_col, last_row, last_col)
Returns
-------
img_array : ndarray
The image data.
Raises
------
RuntimeError
If the image has differing subsample factors.
"""
self._subsampling_sanity_check()
self._populate_dparams(rlevel)
with ExitStack() as stack:
try:
self._dparams.decod_format = self._codec_format
dinfo = opj.create_decompress(self._dparams.decod_format)
event_mgr = opj.EventMgrType()
info_handler = ctypes.cast(_INFO_CALLBACK, ctypes.c_void_p)
if verbose or self._verbose:
event_mgr.info_handler = info_handler
else:
event_mgr.info_handler = None
event_mgr.warning_handler = ctypes.cast(_WARNING_CALLBACK,
ctypes.c_void_p)
event_mgr.error_handler = ctypes.cast(_ERROR_CALLBACK,
ctypes.c_void_p)
opj.set_event_mgr(dinfo, ctypes.byref(event_mgr))
opj.setup_decoder(dinfo, self._dparams)
with open(self.filename, 'rb') as fptr:
src = fptr.read()
cio = opj.cio_open(dinfo, src)
image = opj.decode(dinfo, cio)
stack.callback(opj.image_destroy, image)
stack.callback(opj.destroy_decompress, dinfo)
stack.callback(opj.cio_close, cio)
data = extract_image_cube(image)
except ValueError:
opj2.check_error(0)
if data.shape[2] == 1:
# The third dimension has just a single layer. Make the image
# data 2D instead of 3D.
data.shape = data.shape[0:2]
if area is not None:
x0, y0, x1, y1 = area
extent = 2 ** rlevel
if x1 - x0 < extent or y1 - y0 < extent:
msg = "Decoded area is too small."
raise IOError(msg)
area = [int(round(float(x)/extent + 2 ** -20)) for x in area]
rows = slice(area[0], area[2], None)
cols = slice(area[1], area[3], None)
data = data[rows, cols]
return data
def _read_openjp2(self, rlevel=0, layer=None, area=None, tile=None,
verbose=False):
"""Read a JPEG 2000 image using libopenjp2.
Parameters
----------
layer : int, optional
Number of quality layer to decode.
rlevel : int, optional
Factor by which to rlevel output resolution. Use -1 to get the
lowest resolution thumbnail.
area : tuple, optional
Specifies decoding image area,
(first_row, first_col, last_row, last_col)
tile : int, optional
Number of tile to decode.
verbose : bool, optional
Print informational messages produced by the OpenJPEG library.
Returns
-------
img_array : ndarray
The image data.
Raises
------
RuntimeError
If the image has differing subsample factors.
"""
if layer is not None:
self._layer = layer
self._subsampling_sanity_check()
self._populate_dparams(rlevel, tile=tile, area=area)
with ExitStack() as stack:
if re.match("2.1", version.openjpeg_version):
filename = self.filename
stream = opj2.stream_create_default_file_stream(filename, True)
stack.callback(opj2.stream_destroy, stream)
else:
fptr = libc.fopen(self.filename, 'rb')
stack.callback(libc.fclose, fptr)
stream = opj2.stream_create_default_file_stream(fptr, True)
stack.callback(opj2.stream_destroy, stream)
codec = opj2.create_decompress(self._codec_format)
stack.callback(opj2.destroy_codec, codec)
opj2.set_error_handler(codec, _ERROR_CALLBACK)
opj2.set_warning_handler(codec, _WARNING_CALLBACK)
if self._verbose or verbose:
opj2.set_info_handler(codec, _INFO_CALLBACK)
else:
opj2.set_info_handler(codec, None)
opj2.setup_decoder(codec, self._dparams)
image = opj2.read_header(stream, codec)
stack.callback(opj2.image_destroy, image)
if self._dparams.nb_tile_to_decode:
opj2.get_decoded_tile(codec, stream, image,
self._dparams.tile_index)
else:
opj2.set_decode_area(codec, image,
self._dparams.DA_x0, self._dparams.DA_y0,
self._dparams.DA_x1, self._dparams.DA_y1)
opj2.decode(codec, stream, image)
opj2.end_decompress(codec, stream)
img_array = extract_image_cube(image)
if img_array.shape[2] == 1:
img_array.shape = img_array.shape[0:2]
return img_array
def _populate_dparams(self, rlevel, tile=None, area=None):
"""Populate decompression structure with appropriate input parameters.
Parameters
----------
rlevel : int
Factor by which to rlevel output resolution.
area : tuple
Specifies decoding image area,
(first_row, first_col, last_row, last_col)
tile : int
Number of tile to decode.
"""
if opj2.OPENJP2 is not None:
dparam = opj2.set_default_decoder_parameters()
else:
dparam = opj.DecompressionParametersType()
opj.set_default_decoder_parameters(ctypes.byref(dparam))
infile = self.filename.encode()
nelts = opj2.PATH_LEN - len(infile)
infile += b'0' * nelts
dparam.infile = infile
if self.ignore_pclr_cmap_cdef:
# Return raw codestream components.
dparam.flags |= 1
dparam.decod_format = self._codec_format
dparam.cp_layer = self._layer
# Must check the specified rlevel against the maximum.
if rlevel != 0:
# Must check the specified rlevel against the maximum.
max_rlevel = self.codestream.segment[2].spcod[4]
if rlevel == -1:
# -1 is shorthand for the largest rlevel
rlevel = max_rlevel
elif rlevel < -1 or rlevel > max_rlevel:
msg = "rlevel must be in the range [-1, {0}] for this image."
msg = msg.format(max_rlevel)
raise IOError(msg)
dparam.cp_reduce = rlevel
if area is not None:
if area[0] < 0 or area[1] < 0 or area[2] <= 0 or area[3] <= 0:
msg = "Upper left corner coordinates must be nonnegative and "
msg += "lower right corner coordinates must be positive: {0}"
raise IOError(msg.format(area))
dparam.DA_y0 = area[0]
dparam.DA_x0 = area[1]
dparam.DA_y1 = area[2]
dparam.DA_x1 = area[3]
if tile is not None:
dparam.tile_index = tile
dparam.nb_tile_to_decode = 1
if self.ignore_pclr_cmap_cdef:
# Return raw codestream components.
dparam.flags |= 1
self._dparams = dparam
def read_bands(self, rlevel=0, layer=None, area=None, tile=None,
verbose=False, ignore_pclr_cmap_cdef=False):
"""Read a JPEG 2000 image.
The only time you should use this method is when the image has
different subsampling factors across components. Otherwise you should
use the read method.
Parameters
----------
layer : int, optional
Number of quality layer to decode.
rlevel : int, optional
Factor by which to rlevel output resolution.
area : tuple, optional
Specifies decoding image area,
(first_row, first_col, last_row, last_col)
tile : int, optional
Number of tile to decode.
ignore_pclr_cmap_cdef : bool
Whether or not to ignore the pclr, cmap, or cdef boxes during any
color transformation. Defaults to False.
verbose : bool, optional
Print informational messages produced by the OpenJPEG library.
Returns
-------
lst : list
The individual image components.
See also
--------
read : read JPEG 2000 image
Examples
--------
>>> import glymur
>>> jfile = glymur.data.nemo()
>>> jp = glymur.Jp2k(jfile)
>>> components_lst = jp.read_bands(rlevel=1)
"""
if version.openjpeg_version_tuple[0] < 2:
raise RuntimeError("You must have at least version 2.0.0 of "
"OpenJPEG installed before using this "
"functionality.")
self.ignore_pclr_cmap_cdef = ignore_pclr_cmap_cdef
if layer is not None:
self._layer = layer
self._populate_dparams(rlevel, tile=tile, area=area)
with ExitStack() as stack:
if re.match("2.1", version.openjpeg_version):
# API change in 2.1
filename = self.filename
stream = opj2.stream_create_default_file_stream(filename, True)
stack.callback(opj2.stream_destroy, stream)
else:
fptr = libc.fopen(self.filename, 'rb')
stack.callback(libc.fclose, fptr)
stream = opj2.stream_create_default_file_stream(fptr, True)
stack.callback(opj2.stream_destroy, stream)
codec = opj2.create_decompress(self._codec_format)
stack.callback(opj2.destroy_codec, codec)
opj2.set_error_handler(codec, _ERROR_CALLBACK)
opj2.set_warning_handler(codec, _WARNING_CALLBACK)
if verbose:
opj2.set_info_handler(codec, _INFO_CALLBACK)
else:
opj2.set_info_handler(codec, None)
opj2.setup_decoder(codec, self._dparams)
image = opj2.read_header(stream, codec)
stack.callback(opj2.image_destroy, image)
if self._dparams.nb_tile_to_decode:
opj2.get_decoded_tile(codec, stream, image,
self._dparams.tile_index)
else:
opj2.set_decode_area(codec, image,
self._dparams.DA_x0, self._dparams.DA_y0,
self._dparams.DA_x1, self._dparams.DA_y1)
opj2.decode(codec, stream, image)
opj2.end_decompress(codec, stream)
lst = extract_image_bands(image)
return lst
def get_codestream(self, header_only=True):
"""Returns a codestream object.
Parameters
----------
header_only : bool, optional
If True, only marker segments in the main header are parsed.
Supplying False may impose a large performance penalty.
Returns
-------
Object describing the codestream syntax.
Examples
--------
>>> import glymur
>>> jfile = glymur.data.nemo()
>>> jp2 = glymur.Jp2k(jfile)
>>> codestream = jp2.get_codestream()
>>> print(codestream.segment[1])
SIZ marker segment @ (3233, 47)
Profile: no profile
Reference Grid Height, Width: (1456 x 2592)
Vertical, Horizontal Reference Grid Offset: (0 x 0)
Reference Tile Height, Width: (1456 x 2592)
Vertical, Horizontal Reference Tile Offset: (0 x 0)
Bitdepth: (8, 8, 8)
Signed: (False, False, False)
Vertical, Horizontal Subsampling: ((1, 1), (1, 1), (1, 1))
"""
with open(self.filename, 'rb') as fptr:
if self._codec_format == opj2.CODEC_J2K:
codestream = Codestream(fptr, self.length,
header_only=header_only)
else:
box = [x for x in self.box if x.box_id == 'jp2c']
fptr.seek(box[0].offset)
read_buffer = fptr.read(8)
(box_length, _) = struct.unpack('>I4s', read_buffer)
if box_length == 0:
# The length of the box is presumed to last until the end
# of the file. Compute the effective length of the box.
box_length = os.path.getsize(fptr.name) - fptr.tell() + 8
elif box_length == 1:
# Seek past the XL field.
read_buffer = fptr.read(8)
box_length, = struct.unpack('>Q', read_buffer)
codestream = Codestream(fptr, box_length - 8,
header_only=header_only)
return codestream
def _populate_image_struct(self, image, imgdata):
"""Populates image struct needed for compression.
Parameters
----------
image : ImageType(ctypes.Structure)
Corresponds to image_t type in openjp2 headers.
img_array : ndarray
Image data to be written to file.
"""
numrows, numcols, num_comps = imgdata.shape
# set image offset and reference grid
image.contents.x0 = self._cparams.image_offset_x0
image.contents.y0 = self._cparams.image_offset_y0
image.contents.x1 = (image.contents.x0 +
(numcols - 1) * self._cparams.subsampling_dx + 1)
image.contents.y1 = (image.contents.y0 +
(numrows - 1) * self._cparams.subsampling_dy + 1)
# Stage the image data to the openjpeg data structure.
for k in range(0, num_comps):
if re.match("2.0", version.openjpeg_version) is not None:
# 2.0 API
if self._cparams.cp_cinema:
image.contents.comps[k].prec = 12
image.contents.comps[k].bpp = 12
else:
# 2.1 API
if self._cparams.rsiz in (core.OPJ_PROFILE_CINEMA_2K,
core.OPJ_PROFILE_CINEMA_4K):
image.contents.comps[k].prec = 12
image.contents.comps[k].bpp = 12
layer = np.ascontiguousarray(imgdata[:, :, k], dtype=np.int32)
dest = image.contents.comps[k].data
src = layer.ctypes.data
ctypes.memmove(dest, src, layer.nbytes)
return image
def _populate_comptparms(self, img_array):
"""Instantiate and populate comptparms structure.
This structure defines the image components.
Parameters
----------
img_array : ndarray
Image data to be written to file.
"""
# Only two precisions are possible.
if img_array.dtype == np.uint8:
comp_prec = 8
else:
comp_prec = 16
numrows, numcols, num_comps = img_array.shape
if version.openjpeg_version_tuple[0] == 1:
comptparms = (opj.ImageComptParmType * num_comps)()
else:
comptparms = (opj2.ImageComptParmType * num_comps)()
for j in range(num_comps):
comptparms[j].dx = self._cparams.subsampling_dx
comptparms[j].dy = self._cparams.subsampling_dy
comptparms[j].w = numcols
comptparms[j].h = numrows
comptparms[j].x0 = self._cparams.image_offset_x0
comptparms[j].y0 = self._cparams.image_offset_y0
comptparms[j].prec = comp_prec
comptparms[j].bpp = comp_prec
comptparms[j].sgnd = 0
self._comptparms = comptparms
def _component2dtype(component):
"""Take an OpenJPEG component structure and determine the numpy datatype.
Parameters
----------
component : ctypes pointer to ImageCompType (image_comp_t)
single image component structure.
Returns
-------
dtype : builtins.type
numpy datatype to be used to construct an image array.
"""
if component.sgnd:
if component.prec <= 8:
dtype = np.int8
elif component.prec <= 16:
dtype = np.int16
else:
raise RuntimeError("Unhandled precision, datatype")
else:
if component.prec <= 8:
dtype = np.uint8
elif component.prec <= 16:
dtype = np.uint16
else:
raise RuntimeError("Unhandled precision, datatype")
return dtype
def _validate_nonzero_image_size(nrows, ncols, component_index):
"""The image cannot have area of zero.
"""
if nrows == 0 or ncols == 0:
# Letting this situation continue would segfault Python.
msg = "Component {0} has dimensions {1} x {2}"
msg = msg.format(component_index, nrows, ncols)
raise IOError(msg)
JP2_IDS = ['colr', 'cdef', 'cmap', 'jp2c', 'ftyp', 'ihdr', 'jp2h', 'jP ',
'pclr', 'res ', 'resc', 'resd', 'xml ', 'ulst', 'uinf', 'url ',
'uuid']
def _validate_jp2_box_sequence(boxes):
"""Run through series of tests for JP2 box legality.
This is non-exhaustive.
"""
_validate_signature_compatibility(boxes)
_validate_jp2h(boxes)
_validate_jp2c(boxes)
if boxes[1].brand == 'jpx ':
_validate_jpx_box_sequence(boxes)
else:
# Validate the JP2 box IDs.
count = _collect_box_count(boxes)
for box_id in count.keys():
if box_id not in JP2_IDS:
msg = "The presence of a '{0}' box requires that the file "
msg += "type brand be set to 'jpx '."
raise IOError(msg.format(box_id))
_validate_jp2_colr(boxes)
def _validate_jp2_colr(boxes):
"""
Validate JP2 requirements on colour specification boxes.
"""
lst = [box for box in boxes if box.box_id == 'jp2h']
jp2h = lst[0]
for colr in [box for box in jp2h.box if box.box_id == 'colr']:
if colr.approximation != 0:
msg = "A JP2 colr box cannot have a non-zero approximation field."
raise IOError(msg)
def _validate_jpx_box_sequence(boxes):
"""Run through series of tests for JPX box legality."""
_validate_label(boxes)
_validate_jpx_brand(boxes, boxes[1].brand)
_validate_jpx_compatibility(boxes, boxes[1].compatibility_list)
_validate_singletons(boxes)
_validate_top_level(boxes)
def _validate_signature_compatibility(boxes):
"""Validate the file signature and compatibility status."""
# Check for a bad sequence of boxes.
# 1st two boxes must be 'jP ' and 'ftyp'
if boxes[0].box_id != 'jP ' or boxes[1].box_id != 'ftyp':
msg = "The first box must be the signature box and the second "
msg += "must be the file type box."
raise IOError(msg)
# The compatibility list must contain at a minimum 'jp2 '.
if 'jp2 ' not in boxes[1].compatibility_list:
msg = "The ftyp box must contain 'jp2 ' in the compatibility list."
raise IOError(msg)
def _validate_jp2c(boxes):
"""Validate the codestream box in relation to other boxes."""
# jp2c must be preceeded by jp2h
jp2h_lst = [idx for (idx, box) in enumerate(boxes)
if box.box_id == 'jp2h']
jp2h_idx = jp2h_lst[0]
jp2c_lst = [idx for (idx, box) in enumerate(boxes)
if box.box_id == 'jp2c']
if len(jp2c_lst) == 0:
msg = "A codestream box must be defined in the outermost "
msg += "list of boxes."
raise IOError(msg)
jp2c_idx = jp2c_lst[0]
if jp2h_idx >= jp2c_idx:
msg = "The codestream box must be preceeded by a jp2 header box."
raise IOError(msg)
def _validate_jp2h(boxes):
"""Validate the JP2 Header box."""
_check_jp2h_child_boxes(boxes, 'top-level')
jp2h_lst = [box for box in boxes if box.box_id == 'jp2h']
jp2h = jp2h_lst[0]
# 1st jp2 header box cannot be empty.
if len(jp2h.box) == 0:
msg = "The JP2 header superbox cannot be empty."
raise IOError(msg)
# 1st jp2 header box must be ihdr
if jp2h.box[0].box_id != 'ihdr':
msg = "The first box in the jp2 header box must be the image "
msg += "header box."
raise IOError(msg)
# colr must be present in jp2 header box.
colr_lst = [j for (j, box) in enumerate(jp2h.box)
if box.box_id == 'colr']
if len(colr_lst) == 0:
msg = "The jp2 header box must contain a color definition box."
raise IOError(msg)
colr = jp2h.box[colr_lst[0]]
_validate_channel_definition(jp2h, colr)
def _validate_channel_definition(jp2h, colr):
"""Validate the channel definition box."""
cdef_lst = [j for (j, box) in enumerate(jp2h.box) if box.box_id == 'cdef']
if len(cdef_lst) > 1:
msg = "Only one channel definition box is allowed in the "
msg += "JP2 header."
raise IOError(msg)
elif len(cdef_lst) == 1:
cdef = jp2h.box[cdef_lst[0]]
if colr.colorspace == core.SRGB:
if any([chan + 1 not in cdef.association
or cdef.channel_type[chan] != 0
for chan in [0, 1, 2]]):
msg = "All color channels must be defined in the "
msg += "channel definition box."
raise IOError(msg)
elif colr.colorspace == core.GREYSCALE:
if 0 not in cdef.channel_type:
msg = "All color channels must be defined in the "
msg += "channel definition box."
raise IOError(msg)
JP2H_CHILDREN = set(['bpcc', 'cdef', 'cmap', 'ihdr', 'pclr'])
def _check_jp2h_child_boxes(boxes, parent_box_name):
"""Certain boxes can only reside in the JP2 header."""
box_ids = set([box.box_id for box in boxes])
intersection = box_ids.intersection(JP2H_CHILDREN)
if len(intersection) > 0 and parent_box_name not in ['jp2h', 'jpch']:
msg = "A '{0}' box can only be nested in a JP2 header box."
raise IOError(msg.format(list(intersection)[0]))
# Recursively check any contained superboxes.
for box in boxes:
if hasattr(box, 'box'):
_check_jp2h_child_boxes(box.box, box.box_id)
def _collect_box_count(boxes):
"""Count the occurences of each box type."""
count = Counter([box.box_id for box in boxes])
# Add the counts in the superboxes.
for box in boxes:
if hasattr(box, 'box'):
count.update(_collect_box_count(box.box))
return count
TOP_LEVEL_ONLY_BOXES = set(['dtbl'])
def _check_superbox_for_top_levels(boxes):
"""Several boxes can only occur at the top level."""
# We are only looking at the boxes contained in a superbox, so if any of
# the blacklisted boxes show up here, it's an error.
box_ids = set([box.box_id for box in boxes])
intersection = box_ids.intersection(TOP_LEVEL_ONLY_BOXES)
if len(intersection) > 0:
msg = "A '{0}' box cannot be nested in a superbox."
raise IOError(msg.format(list(intersection)[0]))
# Recursively check any contained superboxes.
for box in boxes:
if hasattr(box, 'box'):
_check_superbox_for_top_levels(box.box)
def _validate_top_level(boxes):
"""Several boxes can only occur at the top level."""
# Add the counts in the superboxes.
for box in boxes:
if hasattr(box, 'box'):
_check_superbox_for_top_levels(box.box)
count = _collect_box_count(boxes)
# Which boxes occur more than once?
multiples = [box_id for box_id, bcount in count.items() if bcount > 1]
if 'dtbl' in multiples:
raise IOError('There can only be one dtbl box in a file.')
# If there is one data reference box, then there must also be one ftbl.
if 'dtbl' in count and 'ftbl' not in count:
msg = 'The presence of a data reference box requires the presence of '
msg += 'a fragment table box as well.'
raise IOError(msg)
def _validate_singletons(boxes):
"""Several boxes can only occur once."""
count = _collect_box_count(boxes)
# Which boxes occur more than once?
multiples = [box_id for box_id, bcount in count.items() if bcount > 1]
if 'dtbl' in multiples:
raise IOError('There can only be one dtbl box in a file.')
JPX_IDS = ['asoc', 'nlst']
def _validate_jpx_brand(boxes, brand):
"""
If there is a JPX box then the brand must be 'jpx '.
"""
for box in boxes:
if box.box_id in JPX_IDS:
if brand != 'jpx ':
msg = "A JPX box requires that the file type box brand be "
msg += "'jpx '."
raise RuntimeError(msg)
if hasattr(box, 'box') != 0:
# Same set of checks on any child boxes.
_validate_jpx_brand(box.box, brand)
def _validate_jpx_compatibility(boxes, compatibility_list):
"""
If there is a JPX box then the compatibility list must also contain 'jpx '.
"""
jpx_cl = set(compatibility_list)
for box in boxes:
if box.box_id in JPX_IDS:
if len(set(['jpx ', 'jpxb']).intersection(jpx_cl)) == 0:
msg = "A JPX box requires that either 'jpx ' or 'jpxb' be "
msg += "present in the ftype compatibility list."
raise RuntimeError(msg)
if hasattr(box, 'box') != 0:
# Same set of checks on any child boxes.
_validate_jpx_compatibility(box.box, compatibility_list)
def _validate_label(boxes):
"""
Label boxes can only be inside association, codestream headers, or
compositing layer header boxes.
"""
for box in boxes:
if box.box_id != 'asoc':
if hasattr(box, 'box'):
for boxi in box.box:
if boxi.box_id == 'lbl ':
msg = "A label box cannot be nested inside a {0} box."
msg = msg.format(box.box_id)
raise IOError(msg)
# Same set of checks on any child boxes.
_validate_label(box.box)
def extract_image_cube(image):
"""Extract 3D image from openjpeg data structure.
"""
ncomps = image.contents.numcomps
component = image.contents.comps[0]
dtype = _component2dtype(component)
nrows = component.h
ncols = component.w
data = np.zeros((nrows, ncols, ncomps), dtype)
for k in range(image.contents.numcomps):
component = image.contents.comps[k]
nrows = component.h
ncols = component.w
_validate_nonzero_image_size(nrows, ncols, k)
addr = ctypes.addressof(component.data.contents)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
nelts = nrows * ncols
band = np.ctypeslib.as_array(
(ctypes.c_int32 * nelts).from_address(addr))
data[:, :, k] = np.reshape(band.astype(dtype), (nrows, ncols))
return data
def extract_image_bands(image):
"""Extract unequally-sized image bands.
This routine need only be called when subsampling differs across image
components, such as is often the case with YCbCr imagery.
"""
data = []
for k in range(image.contents.numcomps):
component = image.contents.comps[k]
dtype = _component2dtype(component)
nrows = component.h
ncols = component.w
_validate_nonzero_image_size(nrows, ncols, k)
addr = ctypes.addressof(component.data.contents)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
band = np.ctypeslib.as_array(
(ctypes.c_int32 * nrows * ncols).from_address(addr))
data.append(np.reshape(band.astype(dtype), (nrows, ncols)))
return data
# Setup the default callback handlers. See the callback functions subsection
# in the ctypes section of the Python documentation for a solid explanation of
# what's going on here.
_CMPFUNC = ctypes.CFUNCTYPE(ctypes.c_void_p, ctypes.c_char_p, ctypes.c_void_p)
def _default_error_handler(msg, _):
"""Default error handler callback for libopenjp2."""
msg = "OpenJPEG library error: {0}".format(msg.decode('utf-8').rstrip())
opj2.set_error_message(msg)
def _default_info_handler(msg, _):
"""Default info handler callback."""
print("[INFO] {0}".format(msg.decode('utf-8').rstrip()))
def _default_warning_handler(library_msg, _):
"""Default warning handler callback."""
library_msg = library_msg.decode('utf-8').rstrip()
msg = "OpenJPEG library warning: {0}".format(library_msg)
warnings.warn(msg)
_ERROR_CALLBACK = _CMPFUNC(_default_error_handler)
_INFO_CALLBACK = _CMPFUNC(_default_info_handler)
_WARNING_CALLBACK = _CMPFUNC(_default_warning_handler)