Binning¶

To reduce the time taken for GLCM, we shrink GLCM size by limiting the maximum value of the input.

Note

E.g. an image of max value 255 requires $256 \times 256$ sized GLCM. Binning it to 15 will drastically reduce to $16 \times 16$ .

The arguments used are bin_from=256 and bin_to=16.

>>> from glcm_cupy import GLCM
>>> import numpy as np
>>> from PIL import Image
>>> ar = np.asarray(Image.open("image.jpg"))
>>> g = GLCM(..., bin_from=256, bin_to=16).run(ar)

Warning

As we include 0, if an image’s max value is 255, use bin_from=256 .

Time Scale¶

The time complexity is $O (n^{2})$

E.g. bin_to==a -> bin_to==a * 2, the time needed scales by 4

Recommendations¶

Try bin_to<=16 for testing purposes, then increase when ready to use higher bins.

>>> from glcm_cupy import GLCM, Direction
>>> g = GLCM(bin_from=256, bin_to=16)