{{ message }}
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathImageFilter.py
More file actions
605 lines (494 loc) · 18.3 KB
/
Copy pathImageFilter.py
File metadata and controls
605 lines (494 loc) · 18.3 KB
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
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
#
# The Python Imaging Library.
# $Id$
#
# standard filters
#
# History:
# 1995-11-27 fl Created
# 2002-06-08 fl Added rank and mode filters
# 2003-09-15 fl Fixed rank calculation in rank filter; added expand call
#
# Copyright (c) 1997-2003 by Secret Labs AB.
# Copyright (c) 1995-2002 by Fredrik Lundh.
#
# See the README file for information on usage and redistribution.
#
from __future__ import annotations
import abc
import functools
from collections.abc import Sequence
from types import ModuleType
from typing import TYPE_CHECKING, Any, Callable, cast
if TYPE_CHECKING:
from . import _imaging
from ._typing import NumpyArray
class Filter:
@abc.abstractmethod
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
pass
class MultibandFilter(Filter):
pass
class BuiltinFilter(MultibandFilter):
filterargs: tuple[Any, ...]
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
if image.mode == "P":
msg = "cannot filter palette images"
raise ValueError(msg)
return image.filter(*self.filterargs)
class Kernel(BuiltinFilter):
"""
Create a convolution kernel. This only supports 3x3 and 5x5 integer and floating
point kernels.
Kernels can only be applied to "L" and "RGB" images.
:param size: Kernel size, given as (width, height). This must be (3,3) or (5,5).
:param kernel: A sequence containing kernel weights. The kernel will be flipped
vertically before being applied to the image.
:param scale: Scale factor. If given, the result for each pixel is divided by this
value. The default is the sum of the kernel weights.
:param offset: Offset. If given, this value is added to the result, after it has
been divided by the scale factor.
"""
name = "Kernel"
def __init__(
self,
size: tuple[int, int],
kernel: Sequence[float],
scale: float | None = None,
offset: float = 0,
) -> None:
if scale is None:
# default scale is sum of kernel
scale = functools.reduce(lambda a, b: a + b, kernel)
if size[0] * size[1] != len(kernel):
msg = "not enough coefficients in kernel"
raise ValueError(msg)
self.filterargs = size, scale, offset, kernel
class RankFilter(Filter):
"""
Create a rank filter. The rank filter sorts all pixels in
a window of the given size, and returns the ``rank``'th value.
:param size: The kernel size, in pixels.
:param rank: What pixel value to pick. Use 0 for a min filter,
``size * size / 2`` for a median filter, ``size * size - 1``
for a max filter, etc.
"""
name = "Rank"
def __init__(self, size: int, rank: int) -> None:
self.size = size
self.rank = rank
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
if image.mode == "P":
msg = "cannot filter palette images"
raise ValueError(msg)
image = image.expand(self.size // 2, self.size // 2)
return image.rankfilter(self.size, self.rank)
class MedianFilter(RankFilter):
"""
Create a median filter. Picks the median pixel value in a window with the
given size.
:param size: The kernel size, in pixels.
"""
name = "Median"
def __init__(self, size: int = 3) -> None:
self.size = size
self.rank = size * size // 2
class MinFilter(RankFilter):
"""
Create a min filter. Picks the lowest pixel value in a window with the
given size.
:param size: The kernel size, in pixels.
"""
name = "Min"
def __init__(self, size: int = 3) -> None:
self.size = size
self.rank = 0
class MaxFilter(RankFilter):
"""
Create a max filter. Picks the largest pixel value in a window with the
given size.
:param size: The kernel size, in pixels.
"""
name = "Max"
def __init__(self, size: int = 3) -> None:
self.size = size
self.rank = size * size - 1
class ModeFilter(Filter):
"""
Create a mode filter. Picks the most frequent pixel value in a box with the
given size. Pixel values that occur only once or twice are ignored; if no
pixel value occurs more than twice, the original pixel value is preserved.
:param size: The kernel size, in pixels.
"""
name = "Mode"
def __init__(self, size: int = 3) -> None:
self.size = size
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
return image.modefilter(self.size)
class GaussianBlur(MultibandFilter):
"""Blurs the image with a sequence of extended box filters, which
approximates a Gaussian kernel. For details on accuracy see
<https://www.mia.uni-saarland.de/Publications/gwosdek-ssvm11.pdf>
:param radius: Standard deviation of the Gaussian kernel. Either a sequence of two
numbers for x and y, or a single number for both.
"""
name = "GaussianBlur"
def __init__(self, radius: float | Sequence[float] = 2) -> None:
self.radius = radius
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
xy = self.radius
if isinstance(xy, (int, float)):
xy = (xy, xy)
if xy == (0, 0):
return image.copy()
return image.gaussian_blur(xy)
class BoxBlur(MultibandFilter):
"""Blurs the image by setting each pixel to the average value of the pixels
in a square box extending radius pixels in each direction.
Supports float radius of arbitrary size. Uses an optimized implementation
which runs in linear time relative to the size of the image
for any radius value.
:param radius: Size of the box in a direction. Either a sequence of two numbers for
x and y, or a single number for both.
Radius 0 does not blur, returns an identical image.
Radius 1 takes 1 pixel in each direction, i.e. 9 pixels in total.
"""
name = "BoxBlur"
def __init__(self, radius: float | Sequence[float]) -> None:
xy = radius if isinstance(radius, (tuple, list)) else (radius, radius)
if xy[0] < 0 or xy[1] < 0:
msg = "radius must be >= 0"
raise ValueError(msg)
self.radius = radius
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
xy = self.radius
if isinstance(xy, (int, float)):
xy = (xy, xy)
if xy == (0, 0):
return image.copy()
return image.box_blur(xy)
class UnsharpMask(MultibandFilter):
"""Unsharp mask filter.
See Wikipedia's entry on `digital unsharp masking`_ for an explanation of
the parameters.
:param radius: Blur Radius
:param percent: Unsharp strength, in percent
:param threshold: Threshold controls the minimum brightness change that
will be sharpened
.. _digital unsharp masking: https://en.wikipedia.org/wiki/Unsharp_masking#Digital_unsharp_masking
"""
name = "UnsharpMask"
def __init__(
self, radius: float = 2, percent: int = 150, threshold: int = 3
) -> None:
self.radius = radius
self.percent = percent
self.threshold = threshold
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
return image.unsharp_mask(self.radius, self.percent, self.threshold)
class BLUR(BuiltinFilter):
name = "Blur"
# fmt: off
filterargs = (5, 5), 16, 0, (
1, 1, 1, 1, 1,
1, 0, 0, 0, 1,
1, 0, 0, 0, 1,
1, 0, 0, 0, 1,
1, 1, 1, 1, 1,
)
# fmt: on
class CONTOUR(BuiltinFilter):
name = "Contour"
# fmt: off
filterargs = (3, 3), 1, 255, (
-1, -1, -1,
-1, 8, -1,
-1, -1, -1,
)
# fmt: on
class DETAIL(BuiltinFilter):
name = "Detail"
# fmt: off
filterargs = (3, 3), 6, 0, (
0, -1, 0,
-1, 10, -1,
0, -1, 0,
)
# fmt: on
class EDGE_ENHANCE(BuiltinFilter):
name = "Edge-enhance"
# fmt: off
filterargs = (3, 3), 2, 0, (
-1, -1, -1,
-1, 10, -1,
-1, -1, -1,
)
# fmt: on
class EDGE_ENHANCE_MORE(BuiltinFilter):
name = "Edge-enhance More"
# fmt: off
filterargs = (3, 3), 1, 0, (
-1, -1, -1,
-1, 9, -1,
-1, -1, -1,
)
# fmt: on
class EMBOSS(BuiltinFilter):
name = "Emboss"
# fmt: off
filterargs = (3, 3), 1, 128, (
-1, 0, 0,
0, 1, 0,
0, 0, 0,
)
# fmt: on
class FIND_EDGES(BuiltinFilter):
name = "Find Edges"
# fmt: off
filterargs = (3, 3), 1, 0, (
-1, -1, -1,
-1, 8, -1,
-1, -1, -1,
)
# fmt: on
class SHARPEN(BuiltinFilter):
name = "Sharpen"
# fmt: off
filterargs = (3, 3), 16, 0, (
-2, -2, -2,
-2, 32, -2,
-2, -2, -2,
)
# fmt: on
class SMOOTH(BuiltinFilter):
name = "Smooth"
# fmt: off
filterargs = (3, 3), 13, 0, (
1, 1, 1,
1, 5, 1,
1, 1, 1,
)
# fmt: on
class SMOOTH_MORE(BuiltinFilter):
name = "Smooth More"
# fmt: off
filterargs = (5, 5), 100, 0, (
1, 1, 1, 1, 1,
1, 5, 5, 5, 1,
1, 5, 44, 5, 1,
1, 5, 5, 5, 1,
1, 1, 1, 1, 1,
)
# fmt: on
class Color3DLUT(MultibandFilter):
"""Three-dimensional color lookup table.
Transforms 3-channel pixels using the values of the channels as coordinates
in the 3D lookup table and interpolating the nearest elements.
This method allows you to apply almost any color transformation
in constant time by using pre-calculated decimated tables.
.. versionadded:: 5.2.0
:param size: Size of the table. One int or tuple of (int, int, int).
Minimal size in any dimension is 2, maximum is 65.
:param table: Flat lookup table. A list of ``channels * size**3``
float elements or a list of ``size**3`` channels-sized
tuples with floats. Channels are changed first,
then first dimension, then second, then third.
Value 0.0 corresponds lowest value of output, 1.0 highest.
:param channels: Number of channels in the table. Could be 3 or 4.
Default is 3.
:param target_mode: A mode for the result image. Should have not less
than ``channels`` channels. Default is ``None``,
which means that mode wouldn't be changed.
"""
name = "Color 3D LUT"
def __init__(
self,
size: int | tuple[int, int, int],
table: Sequence[float] | Sequence[Sequence[int]] | NumpyArray,
channels: int = 3,
target_mode: str | None = None,
**kwargs: bool,
) -> None:
if channels not in (3, 4):
msg = "Only 3 or 4 output channels are supported"
raise ValueError(msg)
self.size = size = self._check_size(size)
self.channels = channels
self.mode = target_mode
# Hidden flag `_copy_table=False` could be used to avoid extra copying
# of the table if the table is specially made for the constructor.
copy_table = kwargs.get("_copy_table", True)
items = size[0] * size[1] * size[2]
wrong_size = False
numpy: ModuleType | None = None
if hasattr(table, "shape"):
try:
import numpy
except ImportError:
pass
if numpy and isinstance(table, numpy.ndarray):
numpy_table: NumpyArray = table
if copy_table:
numpy_table = numpy_table.copy()
if numpy_table.shape in [
(items * channels,),
(items, channels),
(size[2], size[1], size[0], channels),
]:
table = numpy_table.reshape(items * channels)
else:
wrong_size = True
else:
if copy_table:
table = list(table)
# Convert to a flat list
if table and isinstance(table[0], (list, tuple)):
raw_table = cast(Sequence[Sequence[int]], table)
flat_table: list[int] = []
for pixel in raw_table:
if len(pixel) != channels:
msg = (
"The elements of the table should "
f"have a length of {channels}."
)
raise ValueError(msg)
flat_table.extend(pixel)
table = flat_table
if wrong_size or len(table) != items * channels:
msg = (
"The table should have either channels * size**3 float items "
"or size**3 items of channels-sized tuples with floats. "
f"Table should be: {channels}x{size[0]}x{size[1]}x{size[2]}. "
f"Actual length: {len(table)}"
)
raise ValueError(msg)
self.table = table
@staticmethod
def _check_size(size: Any) -> tuple[int, int, int]:
try:
_, _, _ = size
except ValueError as e:
msg = "Size should be either an integer or a tuple of three integers."
raise ValueError(msg) from e
except TypeError:
size = (size, size, size)
size = tuple(int(x) for x in size)
for size_1d in size:
if not 2 <= size_1d <= 65:
msg = "Size should be in [2, 65] range."
raise ValueError(msg)
return size
@classmethod
def generate(
cls,
size: int | tuple[int, int, int],
callback: Callable[[float, float, float], tuple[float, ...]],
channels: int = 3,
target_mode: str | None = None,
) -> Color3DLUT:
"""Generates new LUT using provided callback.
:param size: Size of the table. Passed to the constructor.
:param callback: Function with three parameters which correspond
three color channels. Will be called ``size**3``
times with values from 0.0 to 1.0 and should return
a tuple with ``channels`` elements.
:param channels: The number of channels which should return callback.
:param target_mode: Passed to the constructor of the resulting
lookup table.
"""
size_1d, size_2d, size_3d = cls._check_size(size)
if channels not in (3, 4):
msg = "Only 3 or 4 output channels are supported"
raise ValueError(msg)
table: list[float] = [0] * (size_1d * size_2d * size_3d * channels)
idx_out = 0
for b in range(size_3d):
for g in range(size_2d):
for r in range(size_1d):
table[idx_out : idx_out + channels] = callback(
r / (size_1d - 1), g / (size_2d - 1), b / (size_3d - 1)
)
idx_out += channels
return cls(
(size_1d, size_2d, size_3d),
table,
channels=channels,
target_mode=target_mode,
_copy_table=False,
)
def transform(
self,
callback: Callable[..., tuple[float, ...]],
with_normals: bool = False,
channels: int | None = None,
target_mode: str | None = None,
) -> Color3DLUT:
"""Transforms the table values using provided callback and returns
a new LUT with altered values.
:param callback: A function which takes old lookup table values
and returns a new set of values. The number
of arguments which function should take is
``self.channels`` or ``3 + self.channels``
if ``with_normals`` flag is set.
Should return a tuple of ``self.channels`` or
``channels`` elements if it is set.
:param with_normals: If true, ``callback`` will be called with
coordinates in the color cube as the first
three arguments. Otherwise, ``callback``
will be called only with actual color values.
:param channels: The number of channels in the resulting lookup table.
:param target_mode: Passed to the constructor of the resulting
lookup table.
"""
if channels not in (None, 3, 4):
msg = "Only 3 or 4 output channels are supported"
raise ValueError(msg)
ch_in = self.channels
ch_out = channels or ch_in
size_1d, size_2d, size_3d = self.size
table: list[float] = [0] * (size_1d * size_2d * size_3d * ch_out)
idx_in = 0
idx_out = 0
for b in range(size_3d):
for g in range(size_2d):
for r in range(size_1d):
values = self.table[idx_in : idx_in + ch_in]
if with_normals:
values = callback(
r / (size_1d - 1),
g / (size_2d - 1),
b / (size_3d - 1),
*values,
)
else:
values = callback(*values)
table[idx_out : idx_out + ch_out] = values
idx_in += ch_in
idx_out += ch_out
return type(self)(
self.size,
table,
channels=ch_out,
target_mode=target_mode or self.mode,
_copy_table=False,
)
def __repr__(self) -> str:
r = [
f"{self.__class__.__name__} from {self.table.__class__.__name__}",
"size={:d}x{:d}x{:d}".format(*self.size),
f"channels={self.channels:d}",
]
if self.mode:
r.append(f"target_mode={self.mode}")
return "<{}>".format(" ".join(r))
def filter(self, image: _imaging.ImagingCore) -> _imaging.ImagingCore:
from . import Image
return image.color_lut_3d(
self.mode or image.mode,
Image.Resampling.BILINEAR,
self.channels,
self.size[0],
self.size[1],
self.size[2],
self.table,
)
You can’t perform that action at this time.