Mock up of decorator to handle RGB images in grayscale images

adapt_rgb.py

import numpy as np
import matplotlib.pyplot as plt

from skimage import data, color, filter, exposure
from skimage.util import dtype
from scipy.ndimage import gaussian_filter


lab_range = (0, 100)


def adapt_rgb(image_filter):
    def image_filter_adapted(image, *args, **kwargs):
        rgb_behavior = kwargs.pop('rgb_behavior', 'lightness')
        if color.is_rgb(image):
            if rgb_behavior == 'lightness':
                lab = color.rgb2lab(image)
                lightness = lab[:, :, 0]
                lightness = exposure.rescale_intensity(lightness,
                                                       in_range=lab_range,
                                                       out_range=(0, 1))
                lightness = image_filter(lightness)
                in_range = dtype.dtype_range[lightness.dtype.type]

                if np.all(lightness > 0):
                    in_range = (0, in_range[1])
                lightness = exposure.rescale_intensity(lightness,
                                                       in_range=in_range,
                                                       out_range=lab_range)
                lab[:, :, 0] = lightness
                out = color.lab2rgb(lab)
            elif rgb_behavior == 'hsv':
                hsv = color.rgb2hsv(image)
                value = hsv[:, :, 2]
                value = image_filter(value)
                hsv[:, :, 2] = value
                out = color.hsv2rgb(hsv)
            elif rgb_behavior == 'each channel':
                c_new = [image_filter(c) for c in image.T]
                out = np.array(c_new).T
        else:
            out = image_filter(image)
        return out
    return image_filter_adapted


@adapt_rgb
def edges(image):
    return filter.sobel(image)

@adapt_rgb
def smooth(image):
    # return filter.tv_denoise(image) # slow
    return gaussian_filter(image, 10)


fig, axes = plt.subplots(ncols=3, nrows=2)
axes[0, 0].imshow(smooth(data.lena(), rgb_behavior='each channel'))
axes[1, 0].imshow(edges(data.lena(), rgb_behavior='each channel'))
axes[0, 1].imshow(smooth(data.lena(), rgb_behavior='hsv'))
axes[1, 1].imshow(edges(data.lena(), rgb_behavior='hsv'))
axes[0, 2].imshow(smooth(data.lena(), rgb_behavior='lightness'))
axes[1, 2].imshow(edges(data.lena(), rgb_behavior='lightness'))
plt.show()