def template_match(image, path):
template = cv2.imread(path, cv2.COLOR_BGR2GRAY)
template = convert_to_binary_and_invert(template)
if (image.shape[0] < template.shape[0]
or image.shape[1] < template.shape[1]):
return [], 0
img = image.copy()
w, h = template.shape[::-1]
recv2.lines = cv2.matchTemplate(img, template, cv2.TM_CCOEFF_NORMED)
threshold = 0.65
loc = np.where(res >= threshold)
# print("template width", template.shape[1])
points = []
for pt in zip(*loc[::-1]):
if (len(points) > 0):
if (pt[0] - points[-1] < template.shape[1]):
print("here")
continue
# cv2.line(img, (pt[0], 0), (pt[0], img.shape[0]), (255, 255, 255), 1)
# cv2.line(img, (pt[0] + template.shape[1], 0), (pt[0] + template.shape[1], img.shape[0]), (255, 255, 255), 1) # noqa
# cv2.rectangle(img, pt, (pt[0] + w, pt[1] + h), (255, 255, 255), 2)
points.append(pt[0])
# display_image('res.png', img)
return points, template.shape[1]
def process_image(line_segmets_path, input_path, f, acc_char_map, train):
image = cv2.imread(os.path.join(input_path, f))
# display_image("source", image)
start_time = time.time()
processed_image = convert_to_binary_and_invert(image)
processed_image = deskew(processed_image)
# display_image("after deskew", processed_image)
# cv2.imwrite("binary.png", processed_image)
line_segmets_path = os.path.join(line_segmets_path, f[:-4])
lines = segment_lines(processed_image, line_segmets_path, 0)
curr_ww, curr_tw, acc_char_map, end_time = segment_words(
lines, line_segmets_path, f, input_path, train, acc_char_map) # noqa
if (train):
print(f'we got {curr_ww} wrong out of {curr_tw}')
return curr_ww, curr_tw, acc_char_map, end_time - start_time
default="./segmented_lines")
ap.add_argument("-i",
"--input-path",
required=False,
help="path to line segments file",
default="./inputs")
args = vars(ap.parse_args())
print(args)
input_path = args["input_path"]
line_segmets_path = args["line_segments_path"]
image = cv2.imread("./segmented_lines/csep1622/segment_1/segment_41.png")
display_image("source", image)
processed_image = convert_to_binary_and_invert(image)
# cv2.imwrite("binary.png", processed_image)
image = processed_image.copy() # original copy of the image
img_line = processed_image.copy()
edged = processed_image
# cv2.imwrite("img_cnt.png", edged)
hp = get_horizontal_projection(edged)
baseline = get_baseline_y_coord(get_horizontal_projection(processed_image))
print("now baseline is: ", baseline)
h, w = processed_image.shape
cv2.line(img_line, (0, baseline), (w, baseline), (255, 255, 255), 1)
# cv2.imwrite("baseline.png", img_line)
def segment_words(path):
"""
this function keeps the list of word separatation points in word_separation list
but segments into sub words and saves the sub words segements in their designated directory
"""
files = [
f for f in os.listdir(path) if os.path.isfile(os.path.join(path, f))
]
image = cv2.imread(os.path.join(path, files[0]))
directory_name = path + "/" + files[0][:-4]
if os.path.exists(directory_name):
shutil.rmtree(directory_name)
os.makedirs(directory_name)
original_image = image.copy()
image = convert_to_binary_and_invert(image)
# image_with_line = image.copy()
(h, w) = image.shape
# image_with_line = cv2.dilate(image, np.ones((2, 2), np.uint8), iterations=1) # needs some tuning
horizontal_projection = get_horizontal_projection(image)
baseline_y_coord = get_baseline_y_coord(horizontal_projection)
# cv2.line(image_with_line, (0, baseline_y_coord), (w, baseline_y_coord), (255, 255, 255), 1)
# display_image("image without dotting", image_without_dotting)
vertical_projection = get_vertical_projection(image)
print("shape of vertical projections is: ", len(vertical_projection))
x, count = 0, 0
is_space = False
xcoords = []
distances = []
for i in range(w):
if not is_space:
if vertical_projection[i] == 0:
is_space = True
count = 1
x = i
else:
if vertical_projection[i] > 0:
is_space = False
xcoords.append(x / count)
distances.append(count)
else:
x += i
count += 1
distance = get_distance_between_words(distances)
previous_width = 0
word_separation = xcoords.copy()
print(len(word_separation))
for i in range(len(xcoords)):
if i == 0:
previous_width = int(xcoords[i])
continue
if distances[i - 1] >= distance:
pass
# cv2.line(image, (previous_width, 0), (previous_width, h), (255, 255, 255), 1)
else:
# print(i)
word_separation[i - 1] = -1
sub_word = original_image[:, previous_width:int(xcoords[i])]
cv2.imwrite(directory_name + "/" + "segment_" + str(i) + ".png",
sub_word)
# display_image("sub word", sub_word)
previous_width = int(xcoords[i])
if distances[-2] < distance:
word_separation[-2] = -1
sub_word = original_image[:, int(xcoords[-1]):w]
cv2.imwrite(directory_name + "/" + "segment_" + str(len(xcoords)) + ".png",
sub_word)
display_image("sub word", sub_word)
print("word sep: ", word_separation)
previous_width = 0
sub_seg_points = []
# word_separation = list(filter(lambda a: a != -1, word_separation))
flag = False
for i in range(len(word_separation)):
if word_separation[i] == -1 and flag == False:
flag = True
sub_seg_points = []
sub_seg_points.append(xcoords[i - 1])
# sub_seg_points.append(xcoords[i])
if word_separation[i] == -1 and flag:
sub_seg_points.append(xcoords[i])
if word_separation[i] != -1 and flag:
sub_seg_points.append(xcoords[i])
flag = False
print("sub seg: ", sub_seg_points)
sub_image = image[:,
int(sub_seg_points[0]):int(sub_seg_points[-1])]
display_image("duh", sub_image)
for i in range(1, len(sub_seg_points) - 1):
cv2.line(image, (int(sub_seg_points[i]), 0),
(int(sub_seg_points[i]), h), (255, 255, 255), 1)
display_image("display", image)
previous_width = int(word_separation[i])
cv2.line(image, (int(xcoords[-1]), 0), (int(xcoords[-1]), h),
(255, 255, 255), 1)
print("word: ", word_separation)
print("xcoord: ", xcoords)
display_image("final output", image)
cv2.imwrite("dis.png", image)