def rm_text(org,
corners,
compo_class,
max_text_height=C.THRESHOLD_TEXT_MAX_HEIGHT,
max_text_width=C.THRESHOLD_TEXT_MAX_WIDTH,
ocr_padding=C.OCR_PADDING,
ocr_min_word_area=C.OCR_MIN_WORD_AREA,
show=False):
"""
Remove area that full of text
:param org: original image
:param corners: [(top_left, bottom_right)]
-> top_left: (column_min, row_min)
-> bottom_right: (column_max, row_max)
:param compo_class: classes of corners
:param max_text_height: Too large to be text
:param max_text_width: Too large to be text
:param ocr_padding: Padding for clipping
:param ocr_min_word_area: If too text area ratio is too large
:param show: Show or not
:return: corners without text objects
"""
new_corners = []
new_class = []
for i in range(len(corners)):
corner = corners[i]
(top_left, bottom_right) = corner
(col_min, row_min) = top_left
(col_max, row_max) = bottom_right
height = row_max - row_min
width = col_max - col_min
# highly likely to be block or img if too large
if height > max_text_height and width > max_text_width:
new_corners.append(corner)
new_class.append(compo_class[i])
else:
row_min = row_min - ocr_padding if row_min - ocr_padding >= 0 else 0
row_max = row_max + ocr_padding if row_max + ocr_padding < org.shape[
0] else org.shape[0]
col_min = col_min - ocr_padding if col_min - ocr_padding >= 0 else 0
col_max = col_max + ocr_padding if col_max + ocr_padding < org.shape[
1] else org.shape[1]
# check if this area is text
clip = org[row_min:row_max, col_min:col_max]
if not ocr.is_text(clip, ocr_min_word_area, show=show):
new_corners.append(corner)
new_class.append(compo_class[i])
return new_corners, new_class
def rm_text(org, corners, must_img_height, must_img_width, ocr_padding, ocr_min_word_area, show=False):
new_corners = []
for corner in corners:
(up_left, bottom_right) = corner
(y_min, x_min) = up_left
(y_max, x_max) = bottom_right
height = x_max - x_min
width = y_max - y_min
# highly likely to be block or img if too large
if height > must_img_height and width > must_img_width:
new_corners.append(corner)
else:
# check if this area is text
clip = org[x_min - ocr_padding: x_max + ocr_padding, y_min - ocr_padding: y_max + ocr_padding]
if not ocr.is_text(clip, ocr_min_word_area, show=show):
new_corners.append(corner)
return new_corners