def run():
extension = 'bmp'
pic = sys.argv[1] if len(sys.argv) > 1 else 'smiley'
name = 'samples/' + pic
bw_filename = name + '.' + extension
marked_filename = name + '_marked.' + extension
out_filename = name + '_out.' + extension
bw_rgb = misc.imread(bw_filename)
marked_rgb = misc.imread(marked_filename)
bw = color_conv.rgb2yiq(bw_rgb)
marked = color_conv.rgb2yiq(marked_rgb)
Y = np.array(bw[:, :, 0], dtype='float64')
(n, m) = np.shape(bw)[0:2] # extract image dimensions
size = n * m
Wn = 0
if (os.path.isfile(pic + '.mtx')):
Wn = scipy.io.mmread(pic).tocsr()
else:
Wn = build_weights_matrix(Y)
scipy.io.mmwrite(pic, Wn)
## once markes are found
colored = find_marked_locations(bw_rgb, marked_rgb)
## set rows in colored indices
Wn = Wn.tolil()
xy2idx = np.arange(size).reshape(n, m) # [x,y] -> index
for idx in [xy2idx[i, j] for [i, j] in colored]:
Wn[idx] = sparse.csr_matrix(([1.0], ([0], [idx])), shape=(1, size))
LU = scipy.sparse.linalg.splu(Wn.tocsc())
b1 = (marked[:, :, 1]).flatten()
b2 = (marked[:, :, 2]).flatten()
x1 = LU.solve(b1)
x2 = LU.solve(b2)
sol = np.zeros(np.shape(bw_rgb))
sol[:, :, 0] = Y
sol[:, :, 1] = x1.reshape((n, m))
sol[:, :, 2] = x2.reshape((n, m))
sol_rgb = color_conv.yiq2rgb(sol)
misc.imsave(out_filename, sol_rgb)
def run():
extension = 'bmp'
pic = sys.argv[1] if len(sys.argv) > 1 else 'smiley'
name = 'samples/' + pic
bw_filename = name + '.' + extension
marked_filename = name + '_marked.' + extension
out_filename = name + '_out.' + extension
bw_rgb = misc.imread(bw_filename)
marked_rgb = misc.imread(marked_filename)
bw = color_conv.rgb2yiq(bw_rgb)
marked = color_conv.rgb2yiq(marked_rgb)
Y = np.array(bw[:, :, 0], dtype='float64')
(n, m) = np.shape(bw)[0:2] # extract image dimensions
size = n * m
Wn = 0
if (os.path.isfile(pic + '.mtx')):
Wn = scipy.io.mmread(pic).tocsr()
else:
Wn = build_weights_matrix(Y)
scipy.io.mmwrite(pic, Wn)
## once markes are found
colored = find_marked_locations(bw_rgb, marked_rgb)
## set rows in colored indices
Wn = Wn.tolil()
xy2idx = np.arange(size).reshape(n, m) # [x,y] -> index
for idx in [xy2idx[i, j] for [i, j] in colored]:
Wn[idx] = sparse.csr_matrix(([1.0], ([0], [idx])), shape=(1, size))
LU = scipy.sparse.linalg.splu(Wn.tocsc())
b1 = (marked[:, :, 1]).flatten()
b2 = (marked[:, :, 2]).flatten()
x1 = LU.solve(b1)
x2 = LU.solve(b2)
sol = np.zeros(np.shape(bw_rgb))
sol[:, :, 0] = Y
sol[:, :, 1] = x1.reshape((n, m))
sol[:, :, 2] = x2.reshape((n, m))
sol_rgb = color_conv.yiq2rgb(sol)
misc.imsave(out_filename, sol_rgb)
def read_images():
original_image = cv.imread("samples/hair_res.bmp", 0) # open colour image
im = np.zeros((original_image.shape[0], original_image.shape[1], 3))
im[:, :, 0] = original_image
im[:, :, 1] = original_image
im[:, :, 2] = original_image
cv.imwrite("samples/hair_res1.bmp", im)
np.set_printoptions(threshold=np.nan)
segments = slic(im, n_segments=300, sigma=2)
fig = plt.figure("Superpixels -- %d segments" % 300)
ax = fig.add_subplot(1, 1, 1)
print(segments)
ax.imshow(mark_boundaries(im, segments))
plt.axis("off")
plt.show()
feature_vec = np.zeros(
(len(np.unique(segments)), 3)) # creating feature vector
yiq = color_conv.rgb2yiq(im)
for (j, segValue) in enumerate(np.unique(segments)):
y, i, q = cv.split(yiq)
feature_vec[j][0] = np.mean(y[segments == segValue]) # adding r values
feature_vec[j][1] = np.mean(i[segments == segValue]) # adding g values
feature_vec[j][2] = np.mean(q[segments == segValue]) # adding b values
plt.scatter(feature_vec[:, 0], feature_vec[:, 1], label='True Position')
plt.show()
clusters = KMeans(n_clusters=12, random_state=0).fit(feature_vec)
print(len(clusters.labels_))
for i in range(len(clusters.labels_)):
for ii in range(segments.shape[0]):
for ij in range(segments.shape[1]):
if segments[ii, ij] == i:
segments[ii, ij] = clusters.labels_[i]
plt.scatter(feature_vec[:, 0],
feature_vec[:, 1],
c=clusters.labels_,
cmap='rainbow')
print(segments)
fig = plt.figure("Superpixels -- %d segments" % 300)
ax = fig.add_subplot(1, 1, 1)
ax.imshow(mark_boundaries(im, segments))
plt.axis("off")
plt.show()
def run(convert, email, template, alt_template):
bw_filename = convert.source_file.name
marked_filename = convert.marked_file.name
out_filename = bw_filename.replace('source', 'result')
matrix_name = 'matrix/' + bw_filename.replace('.bmp',
'')[bw_filename.rfind('/'):]
bw_rgb = misc.imread(bw_filename)
marked_rgb = misc.imread(marked_filename)
bw = color_conv.rgb2yiq(bw_rgb)
marked = color_conv.rgb2yiq(marked_rgb)
y = np.array(bw[:, :, 0], dtype='float64')
(n, m) = np.shape(bw)[0:2] # extract image dimensions
size = n * m
wn = 0
if os.path.isfile(matrix_name + '.mtx'):
wn = scipy.io.mmread(matrix_name).tocsr()
else:
wn = build_weights_matrix(y)
scipy.io.mmwrite(matrix_name, wn)
## once markes are found
colored = find_marked_locations(bw_rgb, marked_rgb)
## set rows in colored indices
wn = wn.tolil()
xy2idx = np.arange(size).reshape(n, m) # [x,y] -> index
for idx in [xy2idx[i, j] for [i, j] in colored]:
wn[idx] = sparse.csr_matrix(([1.0], ([0], [idx])), shape=(1, size))
lu = scipy.sparse.linalg.splu(wn.tocsc())
b1 = (marked[:, :, 1]).flatten()
b2 = (marked[:, :, 2]).flatten()
x1 = lu.solve(b1)
x2 = lu.solve(b2)
sol = np.zeros(np.shape(bw_rgb))
sol[:, :, 0] = y
sol[:, :, 1] = x1.reshape((n, m))
sol[:, :, 2] = x2.reshape((n, m))
sol_rgb = color_conv.yiq2rgb(sol)
path_name = out_filename[:out_filename.rfind('/')]
if not os.path.exists(path_name):
os.makedirs(path_name)
misc.imsave(out_filename, sol_rgb)
data = dict()
html = render_to_string(template, data)
text = render_to_string(alt_template, data)
msg = EmailMultiAlternatives(subject=u'Преобразование изображения',
body=text,
from_email=settings.DEFAULT_FROM_EMAIL,
to=[email],
alternatives=[(html, 'text/html')])
msg.send()
f = open(out_filename)
convert.destination_file = out_filename #File(f)
convert.sended = True
convert.save()
f.close()