| 1 |
# -*- coding: utf-8 -*- |
| 2 |
|
| 3 |
import numpy as np |
| 4 |
import scipy.ndimage |
| 5 |
|
| 6 |
from ._dispatcher import Dispatcher |
| 7 |
|
| 8 |
__all__ = [ |
| 9 |
"dispatch_convolve", |
| 10 |
"dispatch_correlate", |
| 11 |
"dispatch_laplace", |
| 12 |
"dispatch_prewitt", |
| 13 |
"dispatch_sobel", |
| 14 |
"dispatch_gaussian_filter", |
| 15 |
"dispatch_gaussian_gradient_magnitude", |
| 16 |
"dispatch_gaussian_laplace", |
| 17 |
"dispatch_generic_filter", |
| 18 |
"dispatch_minimum_filter", |
| 19 |
"dispatch_median_filter", |
| 20 |
"dispatch_maximum_filter", |
| 21 |
"dispatch_rank_filter", |
| 22 |
"dispatch_percentile_filter", |
| 23 |
"dispatch_uniform_filter", |
| 24 |
"dispatch_threshold_local_mean", |
| 25 |
] |
| 26 |
|
| 27 |
|
| 28 |
dispatch_convolve = Dispatcher(name="dispatch_convolve") |
| 29 |
dispatch_correlate = Dispatcher(name="dispatch_correlate") |
| 30 |
dispatch_laplace = Dispatcher(name="dispatch_laplace") |
| 31 |
dispatch_prewitt = Dispatcher(name="dispatch_prewitt") |
| 32 |
dispatch_sobel = Dispatcher(name="dispatch_sobel") |
| 33 |
dispatch_gaussian_filter = Dispatcher(name="dispatch_gaussian_filter") |
| 34 |
dispatch_gaussian_gradient_magnitude = Dispatcher(name="dispatch_gaussian_gradient_magnitude") # noqa: E501 |
| 35 |
dispatch_gaussian_laplace = Dispatcher(name="dispatch_gaussian_laplace") |
| 36 |
dispatch_generic_filter = Dispatcher(name="dispatch_generic_filter") |
| 37 |
dispatch_minimum_filter = Dispatcher(name="dispatch_minimum_filter") |
| 38 |
dispatch_median_filter = Dispatcher(name="dispatch_median_filter") |
| 39 |
dispatch_maximum_filter = Dispatcher(name="dispatch_maximum_filter") |
| 40 |
dispatch_rank_filter = Dispatcher(name="dispatch_rank_filter") |
| 41 |
dispatch_percentile_filter = Dispatcher(name="dispatch_percentile_filter") |
| 42 |
dispatch_uniform_filter = Dispatcher(name="dispatch_uniform_filter") |
| 43 |
dispatch_threshold_local_mean = Dispatcher(name="dispatch_threshold_local_mean") # noqa: E501 |
| 44 |
|
| 45 |
|
| 46 |
# ================== convolve ================== |
| 47 |
@dispatch_convolve.register(np.ndarray) |
| 48 |
def numpy_convolve(*args, **kwargs): |
| 49 |
return scipy.ndimage.convolve |
| 50 |
|
| 51 |
|
| 52 |
@dispatch_convolve.register_lazy("cupy") |
| 53 |
def register_cupy_convolve(): |
| 54 |
import cupy |
| 55 |
import cupyx.scipy.ndimage |
| 56 |
|
| 57 |
@dispatch_convolve.register(cupy.ndarray) |
| 58 |
def cupy_convolve(*args, **kwargs): |
| 59 |
return cupyx.scipy.ndimage.convolve |
| 60 |
|
| 61 |
|
| 62 |
# ================== correlate ================== |
| 63 |
@dispatch_correlate.register(np.ndarray) |
| 64 |
def numpy_correlate(*args, **kwargs): |
| 65 |
return scipy.ndimage.correlate |
| 66 |
|
| 67 |
|
| 68 |
@dispatch_correlate.register_lazy("cupy") |
| 69 |
def register_cupy_correlate(): |
| 70 |
import cupy |
| 71 |
import cupyx.scipy.ndimage |
| 72 |
|
| 73 |
@dispatch_correlate.register(cupy.ndarray) |
| 74 |
def cupy_correlate(*args, **kwargs): |
| 75 |
return cupyx.scipy.ndimage.correlate |
| 76 |
|
| 77 |
|
| 78 |
# ================== laplace ================== |
| 79 |
@dispatch_laplace.register(np.ndarray) |
| 80 |
def numpy_laplace(*args, **kwargs): |
| 81 |
return scipy.ndimage.laplace |
| 82 |
|
| 83 |
|
| 84 |
@dispatch_laplace.register_lazy("cupy") |
| 85 |
def register_cupy_laplace(): |
| 86 |
import cupy |
| 87 |
import cupyx.scipy.ndimage |
| 88 |
|
| 89 |
@dispatch_laplace.register(cupy.ndarray) |
| 90 |
def cupy_laplace(*args, **kwargs): |
| 91 |
return cupyx.scipy.ndimage.laplace |
| 92 |
|
| 93 |
|
| 94 |
# ================== prewitt ================== |
| 95 |
@dispatch_prewitt.register(np.ndarray) |
| 96 |
def numpy_prewitt(*args, **kwargs): |
| 97 |
return scipy.ndimage.prewitt |
| 98 |
|
| 99 |
|
| 100 |
@dispatch_prewitt.register_lazy("cupy") |
| 101 |
def register_cupy_prewitt(): |
| 102 |
import cupy |
| 103 |
import cupyx.scipy.ndimage |
| 104 |
|
| 105 |
@dispatch_prewitt.register(cupy.ndarray) |
| 106 |
def cupy_prewitt(*args, **kwargs): |
| 107 |
return cupyx.scipy.ndimage.prewitt |
| 108 |
|
| 109 |
|
| 110 |
# ================== sobel ================== |
| 111 |
@dispatch_sobel.register(np.ndarray) |
| 112 |
def numpy_sobel(*args, **kwargs): |
| 113 |
return scipy.ndimage.sobel |
| 114 |
|
| 115 |
|
| 116 |
@dispatch_sobel.register_lazy("cupy") |
| 117 |
def register_cupy_sobel(): |
| 118 |
import cupy |
| 119 |
import cupyx.scipy.ndimage |
| 120 |
|
| 121 |
@dispatch_sobel.register(cupy.ndarray) |
| 122 |
def cupy_sobel(*args, **kwargs): |
| 123 |
return cupyx.scipy.ndimage.sobel |
| 124 |
|
| 125 |
|
| 126 |
# ================== gaussian_filter ================== |
| 127 |
@dispatch_gaussian_filter.register(np.ndarray) |
| 128 |
def numpy_gaussian_filter(*args, **kwargs): |
| 129 |
return scipy.ndimage.gaussian_filter |
| 130 |
|
| 131 |
|
| 132 |
@dispatch_gaussian_filter.register_lazy("cupy") |
| 133 |
def register_cupy_gaussian_filter(): |
| 134 |
import cupy |
| 135 |
import cupyx.scipy.ndimage |
| 136 |
|
| 137 |
@dispatch_gaussian_filter.register(cupy.ndarray) |
| 138 |
def cupy_gaussian_filter(*args, **kwargs): |
| 139 |
return cupyx.scipy.ndimage.gaussian_filter |
| 140 |
|
| 141 |
|
| 142 |
# ================== gaussian_gradient_magnitude ================== |
| 143 |
@dispatch_gaussian_gradient_magnitude.register(np.ndarray) |
| 144 |
def numpy_gaussian_gradient_magnitude(*args, **kwargs): |
| 145 |
return scipy.ndimage.gaussian_gradient_magnitude |
| 146 |
|
| 147 |
|
| 148 |
@dispatch_gaussian_gradient_magnitude.register_lazy("cupy") |
| 149 |
def register_cupy_gaussian_gradient_magnitude(): |
| 150 |
import cupy |
| 151 |
import cupyx.scipy.ndimage |
| 152 |
|
| 153 |
@dispatch_gaussian_gradient_magnitude.register(cupy.ndarray) |
| 154 |
def cupy_gaussian_gradient_magnitude(*args, **kwargs): |
| 155 |
return cupyx.scipy.ndimage.gaussian_gradient_magnitude |
| 156 |
|
| 157 |
|
| 158 |
# ================== gaussian_laplace ================== |
| 159 |
@dispatch_gaussian_laplace.register(np.ndarray) |
| 160 |
def numpy_gaussian_laplace(*args, **kwargs): |
| 161 |
return scipy.ndimage.gaussian_laplace |
| 162 |
|
| 163 |
|
| 164 |
@dispatch_gaussian_laplace.register_lazy("cupy") |
| 165 |
def register_cupy_gaussian_laplace(): |
| 166 |
import cupy |
| 167 |
import cupyx.scipy.ndimage |
| 168 |
|
| 169 |
@dispatch_gaussian_laplace.register(cupy.ndarray) |
| 170 |
def cupy_gaussian_laplace(*args, **kwargs): |
| 171 |
return cupyx.scipy.ndimage.gaussian_laplace |
| 172 |
|
| 173 |
|
| 174 |
# ================== generic_filter ================== |
| 175 |
@dispatch_generic_filter.register(np.ndarray) |
| 176 |
def numpy_generic_filter(*args, **kwargs): |
| 177 |
return scipy.ndimage.generic_filter |
| 178 |
|
| 179 |
|
| 180 |
@dispatch_generic_filter.register_lazy("cupy") |
| 181 |
def register_cupy_generic_filter(): |
| 182 |
import cupy |
| 183 |
import cupyx.scipy.ndimage |
| 184 |
|
| 185 |
@dispatch_generic_filter.register(cupy.ndarray) |
| 186 |
def cupy_generic_filter(*args, **kwargs): |
| 187 |
return cupyx.scipy.ndimage.generic_filter |
| 188 |
|
| 189 |
|
| 190 |
# ================== minimum_filter ================== |
| 191 |
@dispatch_minimum_filter.register(np.ndarray) |
| 192 |
def numpy_minimum_filter(*args, **kwargs): |
| 193 |
return scipy.ndimage.minimum_filter |
| 194 |
|
| 195 |
|
| 196 |
@dispatch_minimum_filter.register_lazy("cupy") |
| 197 |
def register_cupy_minimum_filter(): |
| 198 |
import cupy |
| 199 |
import cupyx.scipy.ndimage |
| 200 |
|
| 201 |
@dispatch_minimum_filter.register(cupy.ndarray) |
| 202 |
def cupy_minimum_filter(*args, **kwargs): |
| 203 |
return cupyx.scipy.ndimage.minimum_filter |
| 204 |
|
| 205 |
|
| 206 |
# ================== median_filter ================== |
| 207 |
@dispatch_median_filter.register(np.ndarray) |
| 208 |
def numpy_median_filter(*args, **kwargs): |
| 209 |
return scipy.ndimage.median_filter |
| 210 |
|
| 211 |
|
| 212 |
@dispatch_median_filter.register_lazy("cupy") |
| 213 |
def register_cupy_median_filter(): |
| 214 |
import cupy |
| 215 |
import cupyx.scipy.ndimage |
| 216 |
|
| 217 |
@dispatch_median_filter.register(cupy.ndarray) |
| 218 |
def cupy_median_filter(*args, **kwargs): |
| 219 |
return cupyx.scipy.ndimage.median_filter |
| 220 |
|
| 221 |
|
| 222 |
# ================== maximum_filter ================== |
| 223 |
@dispatch_maximum_filter.register(np.ndarray) |
| 224 |
def numpy_maximum_filter(*args, **kwargs): |
| 225 |
return scipy.ndimage.maximum_filter |
| 226 |
|
| 227 |
|
| 228 |
@dispatch_maximum_filter.register_lazy("cupy") |
| 229 |
def register_cupy_maximum_filter(): |
| 230 |
import cupy |
| 231 |
import cupyx.scipy.ndimage |
| 232 |
|
| 233 |
@dispatch_maximum_filter.register(cupy.ndarray) |
| 234 |
def cupy_maximum_filter(*args, **kwargs): |
| 235 |
return cupyx.scipy.ndimage.maximum_filter |
| 236 |
|
| 237 |
|
| 238 |
# ================== rank_filter ================== |
| 239 |
@dispatch_rank_filter.register(np.ndarray) |
| 240 |
def numpy_rank_filter(*args, **kwargs): |
| 241 |
return scipy.ndimage.rank_filter |
| 242 |
|
| 243 |
|
| 244 |
@dispatch_rank_filter.register_lazy("cupy") |
| 245 |
def register_cupy_rank_filter(): |
| 246 |
import cupy |
| 247 |
import cupyx.scipy.ndimage |
| 248 |
|
| 249 |
@dispatch_rank_filter.register(cupy.ndarray) |
| 250 |
def cupy_rank_filter(*args, **kwargs): |
| 251 |
return cupyx.scipy.ndimage.rank_filter |
| 252 |
|
| 253 |
|
| 254 |
# ================== percentile_filter ================== |
| 255 |
@dispatch_percentile_filter.register(np.ndarray) |
| 256 |
def numpy_percentile_filter(*args, **kwargs): |
| 257 |
return scipy.ndimage.percentile_filter |
| 258 |
|
| 259 |
|
| 260 |
@dispatch_percentile_filter.register_lazy("cupy") |
| 261 |
def register_cupy_percentile_filter(): |
| 262 |
import cupy |
| 263 |
import cupyx.scipy.ndimage |
| 264 |
|
| 265 |
@dispatch_percentile_filter.register(cupy.ndarray) |
| 266 |
def cupy_percentile_filter(*args, **kwargs): |
| 267 |
return cupyx.scipy.ndimage.percentile_filter |
| 268 |
|
| 269 |
|
| 270 |
# ================== uniform_filter ================== |
| 271 |
@dispatch_uniform_filter.register(np.ndarray) |
| 272 |
def numpy_uniform_filter(*args, **kwargs): |
| 273 |
return scipy.ndimage.uniform_filter |
| 274 |
|
| 275 |
|
| 276 |
@dispatch_uniform_filter.register_lazy("cupy") |
| 277 |
def register_cupy_uniform_filter(): |
| 278 |
import cupy |
| 279 |
import cupyx.scipy.ndimage |
| 280 |
|
| 281 |
@dispatch_uniform_filter.register(cupy.ndarray) |
| 282 |
def cupy_uniform_filter(*args, **kwargs): |
| 283 |
return cupyx.scipy.ndimage.uniform_filter |
| 284 |
|
| 285 |
|
| 286 |
# ================== threshold_local_mean ================== |
| 287 |
@dispatch_threshold_local_mean.register(np.ndarray) |
| 288 |
def numpy_threshold_local_mean(*args, **kwargs): |
| 289 |
return np.mean |
| 290 |
|
| 291 |
|
| 292 |
@dispatch_threshold_local_mean.register_lazy("cupy") |
| 293 |
def register_cupy_threshold_local_mean(): |
| 294 |
import cupy |
| 295 |
|
| 296 |
@dispatch_threshold_local_mean.register(cupy.ndarray) |
| 297 |
def cupy_threshold_local_mean(*args, **kwargs): |
| 298 |
# Code snippet taken from https://github.com/cupy/cupy/issues/3909 |
| 299 |
my_mean = cupy.ReductionKernel( |
| 300 |
'T x', # input params |
| 301 |
'T y', # output params |
| 302 |
'x', # map |
| 303 |
'a + b', # reduce |
| 304 |
'y = a / _in_ind.size()', # An undocumented variable and a hack |
| 305 |
'0', # identity value |
| 306 |
'mean' # kernel name |
| 307 |
) |
| 308 |
return my_mean |