def interrogate_index_base(img,
gamma=0.2,
alpha=0.10,
tau=2.65,
ltp_threshold=0.2,
nbr_result=1):
# loading the saved index base numpy array
index_base = np.load('index_base.npy', allow_pickle=True)
# list that contains the distances between the query image and the images in the index base
image_distance = []
# applying the pre-processing to the query image
preprocessed_img = pre_processing(img, gamma, alpha, tau)
# getting the image descriptor of the query image
img_descriptor = ltp(preprocessed_img, ltp_threshold)
# calculating the distance between the query image descriptor and the images descriptors in the index base
for i in range(len(index_base)):
image_distance.append([
index_base[i, 0],
distance.euclidean(img_descriptor, index_base[i, 1])
])
# sorting the distances by the smallest distance
ordered_img_distances = sorted(image_distance, key=lambda x: x[1])
# returning the n images with the shortest distance to the index images
return ordered_img_distances[0:nbr_result]
def create_index_base(path,
gamma=0.2,
alpha=0.10,
tau=2.65,
ltp_threshold=0.2):
# listing all the images in the directory
images_list = os.listdir(path)
# initializing a list that contains the index base ([name of the image, image descriptor]
index_base = []
nbr_images = len(images_list)
# for every images in the image base
for i in range(nbr_images):
# reading the image as a gray scale and normalizing it between 0 and 1
img_full_path = path + '/' + images_list[i]
img = cv.imread(img_full_path, cv.IMREAD_GRAYSCALE).astype(
np.float32) / 255.0
# applying the pre-processing to the current image
preprocessed_img = pre_processing(img, gamma, alpha, tau)
# getting the image descriptor for the current image
img_descriptor = ltp_indexation(preprocessed_img, ltp_threshold)
# appending the image name and the descriptor to the index base list
index_base.append([images_list[i], img_descriptor])
# used for printing the progress on console
sys.stdout.write('\r')
sys.stdout.write("[%-20s] %d%%" % ('=' * (i // (nbr_images // 20) + 1),
(i / nbr_images) * 100 + 1))
sys.stdout.flush()
# saving the index base as a numpy array
index_base_np = np.array(index_base)
# calling the garbage collector
gc.collect()
np.save('index_base.npy', index_base_np)
def clean_paragraph(text):
paragraph = []
for s in sentences(text):
paragraph = paragraph + pre_processing(s).split()
return paragraph
def parse_sentences(rdd):
raw = rdd.zipWithIndex().map(swap_kv)
data = raw.flatMap(lambda (_id, text): [(_id, pre_processing(s).split()) for s in sentences(text)])
return data
def interrogate_index_base(img,
gamma=0.2,
alpha=0.10,
tau=2.65,
ltp_threshold=0.2,
ltp_tau=6,
nbr_result=1):
# loading the saved index base numpy array
index_base = np.load('index_base.npy', allow_pickle=True)
# list that contains the distances between the query image and the images in the index base
image_distance = []
# applying the pre-processing to the query image
preprocessed_img = pre_processing(img, gamma, alpha, tau)
# getting the upper and lower images for query image
upper_descriptor, lower_descriptor = ltp_query(preprocessed_img,
ltp_threshold)
# uniform histogram possible values, taken from "https://en.wikipedia.org/wiki/Local_binary_patterns#Concept",
uni_hist_keys = [
0, 1, 2, 3, 4, 6, 7, 8, 12, 14, 15, 16, 24, 28, 30, 31, 32, 48, 56, 60,
62, 63, 64, 96, 112, 120, 124, 126, 127, 128, 129, 131, 135, 143, 159,
191, 192, 193, 195, 199, 207, 223, 224, 225, 227, 231, 239, 240, 241,
243, 247, 248, 249, 251, 252, 253, 254, 255, 'others'
]
# calculating the distance between the query image descriptor and the images descriptors in the index base
for i in range(len(index_base)):
# the sum of distances between the query image and index base images
distance = 0
# dividng the index base into upper base image descriptor and lower base image descriptor
upper_base_img_descriptor = index_base[i, 1][0:59]
lower_base_img_descriptor = index_base[i, 1][59:]
# initializing the upper and lower descriptors
upper_uni_descriptor = dict(
zip(uni_hist_keys, upper_base_img_descriptor))
lower_uni_descriptor = dict(
zip(uni_hist_keys, lower_base_img_descriptor))
# looping through the size of the query image descriptor
for w in range(1, (upper_descriptor.shape[0] - 1)):
for h in range(1, (upper_descriptor.shape[1] - 1)):
# getting the position of the current pixel in the upper and lower query images
pixel_value_upper = upper_descriptor[w, h]
pixel_value_lower = lower_descriptor[w, h]
# checking if the upper or lower current pixel values are uniform
# calculating the distance to the corresponding bin of the pixel value
if pixel_value_upper in uni_hist_keys:
distance += min(
upper_uni_descriptor[pixel_value_upper][w, h], ltp_tau)
else:
distance += min(upper_uni_descriptor['others'][w, h],
ltp_tau)
# the same operations for the lower binary images
if pixel_value_lower in uni_hist_keys:
distance += min(
lower_uni_descriptor[pixel_value_lower][w, h], ltp_tau)
else:
distance += min(lower_uni_descriptor['others'][w, h],
ltp_tau)
# appending the name of the image and the distance to the image distance list
image_distance.append([index_base[i, 0], distance])
# used for printing the progress on console
sys.stdout.write('\r')
sys.stdout.write("[%-20s] %d%%" % ('=' * (i //
(len(index_base) // 20) + 1),
(i / len(index_base)) * 100 + 1))
sys.stdout.flush()
# sorting the distances by the smallest distance
ordered_img_distances = sorted(image_distance, key=lambda x: x[1])
# returning the n images with the shortest distance to the index images
return ordered_img_distances[0:nbr_result]