2023-01-10 10:19:49 +01:00

55 lines
1.7 KiB
Python

import cv2
from cv_analysis.utils.common import normalize_to_gray_scale
def remove_primary_text_regions(image):
"""Removes regions of primary text, meaning no figure descriptions for example, but main text body paragraphs.
Args:
image: Image to remove primary text from.
Returns:
Image with primary text removed.
References:
https://stackoverflow.com/questions/58349726/opencv-how-to-remove-text-from-background
"""
image = apply_threshold_to_image(image)
threshold_image = image.copy()
close_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (17, 7)) # 20,3
close = cv2.morphologyEx(image, cv2.MORPH_CLOSE, close_kernel, iterations=1)
dilate_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (7, 3)) # 5,3
dilate = cv2.dilate(close, dilate_kernel, iterations=1)
cnts, _ = cv2.findContours(dilate, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
cnts = filter(is_likely_primary_text_segment, cnts)
rects = map(cv2.boundingRect, cnts)
image = draw_bboxes(threshold_image, rects)
return image
def apply_threshold_to_image(image):
"""Converts an image to black and white."""
image = normalize_to_gray_scale(image)
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) if len(image.shape) > 2 else image
return cv2.threshold(image, 253, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
def is_likely_primary_text_segment(cnt):
x, y, w, h = cv2.boundingRect(cnt)
return 400 < cv2.contourArea(cnt) < 16000 or w / h > 3
def draw_bboxes(image, bboxes):
for rect in bboxes:
x, y, w, h = rect
cv2.rectangle(image, (x, y), (x + w, y + h), (0, 0, 0), -1)
return image