def test_command_clean_section_with_palette(self):
image_path = fixtures_path('icon.png')
image = cv2.imread(image_path)
palette = utils.get_palette(image[:, :, ::-1], n_colors=2)
clean_image_b64 = self.runner.invoke(
cli.clean,
[image_path, '-c', 2, '-f', 100, '-p',
json.dumps(palette.tolist())]
).output.strip()
clean_image = decode_base64(clean_image_b64)
section_path = fixtures_path('icon_section.png')
section = cv2.imread(section_path)
clean_section_b64 = self.runner.invoke(
cli.clean,
[section_path, '-c', 2, '-f', 100, '-p',
json.dumps(palette.tolist())]
).output.strip()
clean_section = decode_base64(clean_image_b64)
clean_image_colors = sorted(utils.get_color_histogram(
clean_image[:, :, ::-1]).keys())
section_colors = sorted(utils.get_color_histogram(
section[:, :, ::-1]).keys())
clean_section_colors = sorted(utils.get_color_histogram(
clean_section[:, :, ::-1]).keys())
assert 'Error' not in clean_image_b64
assert 'Error' not in clean_section_b64
assert section_colors != clean_section_colors
assert clean_section_colors == clean_image_colors
def test_command_enhance_with_palette(self):
image_path = fixtures_path('icon.png')
image = cv2.imread(image_path)
palette = utils.get_palette(image[:, :, ::-1], n_colors=8)
result = self.runner.invoke(
cli.enhance,
[image_path, '-p', json.dumps(palette.tolist())]
).output.strip()
assert 'Error' not in result
assert len(result) > 0
def test_posterization_linear_4_colors_with_palette(self):
image = self.image
test_image = cv2.imread(fixtures_path('poster_linear4.png'))
palette = utils.get_palette(image.reshape((-1, 3)), 4, method='linear')
# palette is given as RGB and get_color_histogram returns colors
# in the color scheme of image, which by default is BGR
reduce_image = histonets.color_reduction(image,
4,
'linear',
palette=palette)
assert (len(utils.get_color_histogram(test_image)) == len(
utils.get_color_histogram(reduce_image)))
def test_auto_clean_with_palette(self):
image = cv2.imread(fixtures_path('map.png'))
palette = utils.get_palette(image[:, :, ::-1], n_colors=8)
distances = []
for _ in range(10):
clean1 = histonets.auto_clean(image,
colors=8,
sample_fraction=100,
palette=palette)
clean2 = histonets.auto_clean(image, colors=8, sample_fraction=100)
hist1 = utils.get_color_histogram(clean1)
hist1_colors = np.asarray(sorted(hist1.keys()))
hist2 = utils.get_color_histogram(clean2)
hist2_colors = np.asarray(sorted(hist2.keys()))
distances.append(int(np.linalg.norm(hist1_colors - hist2_colors)))
distance = np.array(distances).mean()
# threshold value is experimentally set as the highest distance
# after running the comparison 10k times
threshold = 298 # 976 in L*a*b colorspace
assert distance <= threshold
def test_auto_clean_section_with_palette(self):
image = cv2.imread(fixtures_path('icon.png'))
palette = utils.get_palette(image[:, :, ::-1], n_colors=2)
clean_image = histonets.auto_clean(image,
colors=2,
sample_fraction=100,
palette=palette)
section = cv2.imread(fixtures_path('icon_section.png'))
clean_section = histonets.auto_clean(section,
colors=2,
sample_fraction=100,
palette=palette)
clean_image_colors = sorted(
utils.get_color_histogram(clean_image[:, :, ::-1]).keys())
section_colors = sorted(
utils.get_color_histogram(section[:, :, ::-1]).keys())
clean_section_colors = sorted(
utils.get_color_histogram(clean_section[:, :, ::-1]).keys())
assert section_colors != clean_section_colors
assert clean_section_colors == clean_image_colors