1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370
|
import numpy
from .Qt import QtGui
from . import functions
from .util.cupy_helper import getCupy
from .util.numba_helper import getNumbaFunctions
def _apply_lut_for_uint(xp, image, lut):
# Note: compared to makeARGB(), we have already clipped the data to range
# if lut is 1d, then lut[image] is fastest
# if lut is 2d, then lut.take(image, axis=0) is faster than lut[image]
lut = _convert_2dlut_to_1dlut(xp, lut)
if xp == numpy and (fn_numba := getNumbaFunctions()) is not None:
# numba "take" supports only the 1st 2 arguments of np.take,
# therefore we have to convert the lut to 1d.
# "take" will output a c contiguous array regardless of its input.
image = fn_numba.numba_take(lut, image)
else:
# advanced indexing is memory order aware.
# its output can be either C or F contiguous.
image = lut[image]
if image.dtype == xp.uint32:
# "view" requires c contiguous for numpy < 1.23
image = xp.ascontiguousarray(image)
image = image[..., xp.newaxis].view(xp.uint8)
return image
def _convert_lut_to_rgba(xp, lut):
# converts:
# - None to (256, 4)
# - uint8 (N,) to uint8 (N, 4)
# - uint8 (N, 1) to uint8 (N, 4)
# - uint8 (N, 3) to uint8 (N, 4)
if not (
lut is None
or lut.ndim == 1
or (
lut.ndim == 2
and lut.shape[1] in (1, 3, 4)
)
):
raise ValueError("unsupported lut shape")
N = lut.shape[0] if lut is not None else 256
if lut is None:
lut = xp.arange(N, dtype=xp.uint8)
# convert (N,) to (N, 1)
if lut.ndim == 1:
lut = lut[:, xp.newaxis]
if lut.shape[1] == 4:
return lut
out = xp.full((N, 4), 255, dtype=xp.uint8)
out[:, 0:3] = lut
return out
def _convert_2dlut_to_1dlut(xp, lut):
# converts:
# - uint8 (N, 1) to uint8 (N,)
# - uint8 (N, 3) or (N, 4) to uint32 (N,)
# this allows faster lookup as 1d lookup is faster
if lut.ndim == 1:
return lut
if lut.shape[1] == 3: # rgb
# convert rgb lut to rgba so that it is 32-bits
lut = xp.column_stack([lut, xp.full(lut.shape[0], 255, dtype=xp.uint8)])
if lut.shape[1] == 4: # rgba
lut = lut.view(xp.uint32)
lut = lut.ravel()
return lut
def _rescale_and_lookup_float(xp, image, levels, lut, *, forceApplyLut):
# It is usually more performant to _not_ apply the lut and
# instead use it as an Indexed8 ColorTable. This is only
# applicable if the lut has <= 256 entries.
if forceApplyLut and lut is None:
raise ValueError("forceApplyLut True but lut not provided")
# Decide on maximum scaled value
if lut is not None:
num_colors = lut.shape[0]
max_scale_value = num_colors
else:
num_colors = 256
max_scale_value = 255.0
dtype = xp.min_scalar_type(num_colors - 1)
# note: "dtype == uint16" ==> lut provided ==> mono-channel image
# i.e. multi-channel image ==> lut is None ==> dtype == uint8
#
# the library defaults to using 256-entry luts, so
# "dtype == uint8" is the common case
apply_lut = forceApplyLut or dtype == xp.uint16
minVal, maxVal = levels
rng = maxVal - minVal
rng = 1 if rng == 0 else rng
offset = minVal
scale = max_scale_value / rng
if xp == numpy and (fn_numba := getNumbaFunctions()) is not None:
if apply_lut:
# this path does rescale and apply lut in one step
lut = _convert_2dlut_to_1dlut(xp, lut)
image = fn_numba.rescale_and_lookup(image, scale, offset, lut)
lut = None
if image.dtype == xp.uint32:
# "view" requires c contiguous for numpy < 1.23
image = xp.ascontiguousarray(image)
image = image[..., xp.newaxis].view(xp.uint8)
else:
image = fn_numba.rescale_and_clip(image, scale, offset, 0, num_colors - 1)
else:
image = functions.rescaleData(
image, scale, offset, dtype=dtype, clip=(0, num_colors - 1)
)
if apply_lut:
image = _apply_lut_for_uint(xp, image, lut)
lut = None
# image is now of type uint8
return image, lut
def _combine_levels_and_lut(xp, image, levels, lut):
if (
image.dtype == xp.uint16
and levels is None
and image.ndim == 3
and image.shape[2] == 3
):
# uint16 rgb can't be directly displayed, so make it
# pass through effective lut processing
levels = [0, 65535]
if levels is None and lut is None:
# nothing to combine
return image, lut
# distinguish between lut for levels and colors
levels_lut = None
colors_lut = lut
eflsize = 2 ** (image.itemsize * 8)
if levels is None:
info = xp.iinfo(image.dtype)
minlev, maxlev = info.min, info.max
else:
minlev, maxlev = levels
levdiff = maxlev - minlev
levdiff = 1 if levdiff == 0 else levdiff # don't allow division by 0
offset = minlev
if colors_lut is None:
scale = 255.0 / levdiff
if image.dtype == xp.ubyte and image.ndim == 2:
# uint8 mono image
ind = xp.arange(eflsize)
levels_lut = functions.rescaleData(ind, scale, offset, dtype=xp.ubyte)
# image data is not scaled. instead, levels_lut is used
# as (grayscale) Indexed8 ColorTable to get the same effect.
# due to the small size of the input to rescaleData(), we
# do not bother caching the result
return image, levels_lut
else:
# uint16 mono, uint8 rgb, uint16 rgb
# rescale image data by computation instead of by memory lookup
if xp == numpy and (fn_numba := getNumbaFunctions()) is not None:
image = fn_numba.rescale_and_clip(image, scale, offset, 0, 255)
else:
image = functions.rescaleData(image, scale, offset, dtype=xp.ubyte)
return image, colors_lut
else:
num_colors = colors_lut.shape[0]
scale = num_colors / levdiff
lutdtype = xp.min_scalar_type(num_colors - 1)
if image.dtype == xp.ubyte or lutdtype != xp.ubyte:
# combine if either:
# 1) uint8 mono image
# 2) colors_lut has more entries than will fit within 8-bits
ind = xp.arange(eflsize)
levels_lut = functions.rescaleData(
ind, scale, offset, dtype=lutdtype, clip=(0, num_colors - 1),
)
efflut = colors_lut[levels_lut]
# apply the effective lut early for the following types:
if image.dtype == xp.uint16 and image.ndim == 2:
image = _apply_lut_for_uint(xp, image, efflut)
efflut = None
return image, efflut
else:
# uint16 image with colors_lut <= 256 entries
# don't combine, we will use QImage ColorTable
if xp == numpy and (fn_numba := getNumbaFunctions()) is not None:
image = fn_numba.rescale_and_clip(image, scale, offset, 0, num_colors - 1)
else:
image = functions.rescaleData(
image, scale, offset, dtype=lutdtype, clip=(0, num_colors - 1),
)
return image, colors_lut
def try_make_qimage(image, *, levels, lut, transparentLocations=None):
"""
Internal function to make an QImage from an ndarray without going
through the full generality of makeARGB().
Only certain combinations of input arguments are supported.
"""
# this function assumes that image has no nans.
# checking for nans is an expensive operation; it is expected that
# the caller would want to cache the result rather than have this
# function check for nans unconditionally.
cp = getCupy()
xp = cp.get_array_module(image) if cp else numpy
# float images always need levels
if image.dtype.kind == "f" and levels is None:
return None
if levels is not None:
levels = xp.asarray(levels)
# can't handle multi-channel levels
if levels.ndim != 1:
return None
# if levels is provided, multi-channel images must be 3 channels only.
# (because it doesn't make sense to scale a 4th alpha channel.)
if image.ndim == 3 and image.shape[2] != 3:
return None
if lut is not None and lut.dtype != xp.uint8:
raise ValueError("lut dtype must be uint8")
alpha_channel_required = (
( # image itself has alpha channel
image.ndim == 3
and image.shape[2] == 4
)
or
( # lut has alpha channel
lut is not None
and lut.ndim == 2
and lut.shape[1] == 4
)
)
if image.dtype.kind == "f":
if image.ndim == 2:
# mono float images
if transparentLocations is None:
image, lut = _rescale_and_lookup_float(
xp, image, levels, lut, forceApplyLut=False
)
levels = None
# on return, we will have an uint8 image.
# lut if not None will have <= 256 entries
else:
# this path creates an alpha channel
lut = _convert_lut_to_rgba(xp, lut)
alpha_channel_required = True
image, lut = _rescale_and_lookup_float(
xp, image, levels, lut, forceApplyLut=True
)
levels = None
assert lut is None
image[..., 3][transparentLocations] = 0
else:
# RGB float images
# lut can only be None for RGB images
image, lut = _rescale_and_lookup_float(
xp, image, levels, lut, forceApplyLut=False
)
levels = None
if transparentLocations is not None:
alpha_channel_required = True
mask = xp.full(image.shape[:2], 255, dtype=xp.uint8)
mask[transparentLocations] = 0
image = xp.dstack((image, mask))
# if the image data is a small int, then we can combine levels + lut
# into a single lut for better performance
elif image.dtype in (xp.ubyte, xp.uint16):
image, lut = _combine_levels_and_lut(xp, image, levels, lut)
levels = None
ubyte_nolvl = image.dtype == xp.ubyte and levels is None
is_passthru8 = ubyte_nolvl and lut is None
is_indexed8 = (
ubyte_nolvl and image.ndim == 2 and lut is not None and lut.shape[0] <= 256
)
is_passthru16 = image.dtype == xp.uint16 and levels is None and lut is None
can_grayscale16 = (
is_passthru16
and image.ndim == 2
and hasattr(QtGui.QImage.Format, "Format_Grayscale16")
)
is_rgba64 = is_passthru16 and image.ndim == 3 and image.shape[2] == 4
# bypass makeARGB for supported combinations
supported = is_passthru8 or is_indexed8 or can_grayscale16 or is_rgba64
if not supported:
return None
if xp == cp:
image = image.get()
# worthwhile supporting non-contiguous arrays
image = numpy.ascontiguousarray(image)
fmt = None
ctbl = None
if is_passthru8:
# both levels and lut are None
# these images are suitable for display directly
if image.ndim == 2:
fmt = QtGui.QImage.Format.Format_Grayscale8
elif image.shape[2] == 3:
fmt = QtGui.QImage.Format.Format_RGB888
elif image.shape[2] == 4:
if alpha_channel_required:
fmt = QtGui.QImage.Format.Format_RGBA8888
else:
fmt = QtGui.QImage.Format.Format_RGBX8888
elif is_indexed8:
# levels and/or lut --> lut-only
fmt = QtGui.QImage.Format.Format_Indexed8
if lut.ndim == 1 or lut.shape[1] == 1:
ctbl = [QtGui.qRgb(x, x, x) for x in lut.ravel().tolist()]
elif lut.shape[1] == 3:
ctbl = [QtGui.qRgb(*rgb) for rgb in lut.tolist()]
elif lut.shape[1] == 4:
ctbl = [QtGui.qRgba(*rgba) for rgba in lut.tolist()]
elif can_grayscale16:
# single channel uint16
# both levels and lut are None
fmt = QtGui.QImage.Format.Format_Grayscale16
elif is_rgba64:
# uint16 rgba
# both levels and lut are None
fmt = QtGui.QImage.Format.Format_RGBA64 # endian-independent
if fmt is None:
raise ValueError("unsupported image type")
qimage = functions.ndarray_to_qimage(image, fmt)
if ctbl is not None:
qimage.setColorTable(ctbl)
return qimage
|