def main():
global img_list
global gmap
global weight
img_list = utils.read_images("../../test_data/cafe", N, downsample=3)
full_img_list = utils.read_images("../../test_data/cafe", N)
gray_imgs = utils.read_images("../../test_data/cafe", N, gray=True)
x0 = np.full(N, 1.0/N)
(gmap, weight) = gradient_map(gray_imgs)
bnds = []
for i in range(len(img_list)):
bnds.append((0, 1))
lambdas = minimize(edge_light, x0, method='TNC', jac=False,
bounds=bnds)
ret_image = sum_images(lambdas.x, full_img_list)
print lambdas.message
print "Choice of lambdas = %s" % (lambdas.x)
cv2.imwrite('output_edge.png', utils.denormalize_img(ret_image))
cv2.imshow('image', ret_image)
cv2.waitKey(0)
def diffuse_light(directory, output, count=-1, downsample=0, verbose=False):
"""
This function calculates the fill light of the images.
Arguments:
directory -- the directory where the images are located
count -- the number of imags to use for this calculation
verbose -- should we print debug info
"""
img_list = utils.read_images(directory, count)
d_img_list = utils.read_images(directory, count, downsample=downsample)
basis = basis_lights.BasisLights(img_list, verbose=verbose,
downsampled=d_img_list)
res_image = basis.diffuse_color()
cv2.imwrite(output, utils.denormalize_img(res_image))
cv2.imshow('Diffuse color light', res_image)
cv2.waitKey(0)
def soft_modifier(directory, output, sigma, count=-1, verbose=False):
"""
This function calculates the soft lightning modifier for an image.
Arguments:
directory -- the directory where the images are located
sigma -- determines how far the lights get diffused
count -- the number of imags to use for this calculation
verbose -- should we print debug info
"""
img_list = utils.read_images(directory, count, normalize=False)
modifier = modifier_lights.ModifierLights(img_list, verbose=verbose)
res_image = modifier.soft(sigma)
cv2.imwrite(output, res_image)
cv2.imshow('Soft modifier', res_image)
cv2.waitKey(0)
#path_patches='../DATA_GHANA/RAW_PATCHES/120_x_120/'
path_patches = sys.argv[2]
if not os.path.exists(path_patches):
os.makedirs(path_patches)
for i in range(3, len(sys.argv)):
arg = sys.argv[i]
if arg.startswith('--width_patch'):
WIDTH = int(arg[len('--width_patch='):])
STRIDE = WIDTH
#patch_test_number=300
## Panchromatic
panchromatic_file = path + NAME_PANCHRO
panchromatic = read_images(panchromatic_file)
hr_size = panchromatic.shape
panchromatic = panchromatic[newaxis, :, :, newaxis]
print('\n PANCHROMATIC \n\n')
panchromatic = extract_patches(panchromatic, WIDTH, STRIDE)
## Find patches to discard
keep = np.arange(len(panchromatic))
discard = []
print('Original size of the dataset is %d' % len(panchromatic))
for i in range(len(panchromatic)):
if (np.sum(panchromatic[i, :, :, :])).astype(int) == 0:
discard.append(i)
discard = np.asarray(discard)
keep = np.delete(keep, discard)
#path_patches='../DATA_GHANA/RAW_PATCHES/120_x_120/'
path_patches=sys.argv[2]
if not os.path.exists(path_patches):
os.makedirs(path_patches)
for i in range(3, len(sys.argv)):
arg = sys.argv[i]
if arg.startswith('--width_patch'):
WIDTH=int(arg[len('--width_patch='):])
STRIDE=WIDTH
#patch_test_number=300
## Panchromatic
panchromatic_file=path+NAME_PANCHRO
panchromatic=read_images(panchromatic_file)
hr_size=panchromatic.shape
panchromatic=panchromatic[newaxis,:,:,newaxis]
print('\n PANCHROMATIC \n\n')
panchromatic=extract_patches(panchromatic,WIDTH,STRIDE)
## Find patches to discard
keep=np.arange(len(panchromatic))
discard=[]
print('Original size of the dataset is %d'%len(panchromatic))
for i in range(len(panchromatic)):
if (np.sum(panchromatic[i,:,:,:])).astype(int)==0:
discard.append(i)
discard=np.asarray(discard)
keep=np.delete(keep,discard)
def main():
# px1 = np.array([0, 126, 255], dtype=float)
# px2 = np.array([0, 0, 0], dtype=float)
global img_list
img_list = utils.read_images("../../test_data/cafe", N, downsample=4)
full_img_list = utils.read_images("../../test_data/cafe", N)
# for i in range(0, N):
# # img_name = "../../../input_image/cafe/images/%03d.png" % (i)
# img_name = "../test_data/cafe/%03d.png" % (i)
# img = cv2.imread(img_name)
# img_list.append(utils.normalize_img(img))
# # img_list.append(img)
# print alpha(10, 10)
# print px_avg(400, 500)
# print px_avg(img_list, 0, 0)
# px1 = np.ndarray(shape=(3, 1), dtype=float,
# buffer=np.array([0, 126, 255]))
# px2 = np.ndarray(shape=(1, 3), dtype=float,
# buffer=np.array([[255], [0], [128]]))
# diffuse_color(1)
x0 = np.full(N, 1.0/N)
# print fprime(x0)
# print x0
# print diffuse_color(x0)
# x0 = np.zeros(N)
# print diffuse_color(x1)
# x0 = np.ones(N)
# print diffuse_color(x2)
# results = pyipopt.fmin_unconstrained(diffuse_color, x0, fprime)
# print results
# nlp = pyipopt.create(diffuse_color, N)
# x, zl, zu, constraint_multipliers, obj, status = nlp.solve(x0)
# nlp.close()
# leastsq(diffuse_color, x0)
bnds = []
for i in range(len(img_list)):
bnds.append((0, 1))
lambdas = minimize(diffuse_color, x0, method='TNC', jac=False,
bounds=bnds)
# lambdas = np.array([ 0.0772488 , 0.07700344, 0.0769977 , 0.07699107, 0.07702644,
# 0.07702589, 0.07702531, 0.07694187, 0.07695671, 0.07701653,
# 0.07696243, 0.07767547, 0.07693373])
# lambdas = np.array([ 0.03336186, 0.03335751, 0.03336186, 0.03336186, 0.03335691,
# 0.03336186, 0.03335623, 0.03335545, 0.03336186, 0.03316012,
# 0.03336186, 0.03332896, 0.03336186, 0.03336186, 0.03335172,
# 0.03335502, 0.03336186, 0.03336186, 0.03335585, 0.03333409,
# 0.03335658, 0.03335828, 0.03335722, 0.03335778, 0.03336186,
# 0.03335804, 0.03336186, 0.03335251, 0.03332677, 0.0333048 ])
# print("lambda: %s" % lambdas)
ret_image = sum_images(lambdas.x, full_img_list)
print lambdas.message
print "Choice of lambdas = %s" % (lambdas.x)
cv2.imwrite('output_diffuse.png', utils.denormalize_img(ret_image))
cv2.imshow('image', ret_image)
cv2.waitKey(0)