This python script is written as a part of research for contextual spacing tools to ease extracting letterform shapes from manuscripts

extract-letter-shape-contours-from-image.py

import cv2
import numpy as np
import os
from pathlib import Path
import argparse

"""
Usage:

Install prerequisite python packages:
pip3 install numpy
pip3 install opencv-python

Run the script:
1- cd to the folder that contains the images opened in Mac os Finder:
cd "$(osascript -e 'tell application "Finder" to get the POSIX path of (target of front Finder window as alias)')"
2- python3 'path/to/extract-contours-from-image.py'
"""

expand_2 = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (2, 2))

def extractAndSaveContoursFromImage(imagePath, minImageWidth=50, minImageHeight=50, minImageSize=1000,
    cropmargin=10, expand_contours=3):
    fpo = Path(imagePath)
    fileName = fpo.stem
    fileRoot = fpo.parent

    image = cv2.imread(imagePath)
    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    thresh = cv2.threshold(
        gray, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)[1]
    thresh = cv2.dilate(thresh, expand_2, iterations=1) # connects more shapes that are closer to another (like sarksh kaf)
    output_folder = f"{fileRoot}/contours"
    os.makedirs(output_folder, exist_ok=True)
    output_path = f"{output_folder}/threshold.jpg"
    cv2.imwrite(output_path, thresh) # debug mask
    cnts = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    cnts = cnts[0] if len(cnts) == 2 else cnts[1]
    expand_3 = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (expand_contours, expand_contours))
    for i, c in enumerate(cnts):
        x, y, w, h = cv2.boundingRect(c)
        if w < minImageWidth or h < minImageWidth or w * h < minImageSize:
            continue

        # increase w and h first to account for cropmargin
        h = h + (cropmargin * 2)
        w = w + (cropmargin * 2)
        x = max(x - cropmargin, 0)
        y = max(y - cropmargin, 0)
        cropped = gray[y: y + h, x: x + w]

        mask_image = np.ones((h, w, 3), np.uint8) * 0

        # translate contours by -(x,y) to move it into the top-left direction
        c = c - [x, y]

        # draw contour on mask_image and expand (grow) the contour edge
        cv2.drawContours(mask_image, [c], -1, (255, 255, 255), thickness=cv2.FILLED)
        mask_image = cv2.dilate(mask_image, expand_3, iterations=3) # expands the contours
        mask_image = cv2.GaussianBlur(mask_image, (expand_contours, expand_contours), 0)
        mask_image = cv2.cvtColor(mask_image, cv2.COLOR_BGR2GRAY)
        mask_image = cv2.bitwise_not(mask_image)
        result = cv2.add(mask_image, cropped)
        output_path = os.path.join(output_folder, f'm_{fileName}_{i}.jpg')
        cv2.imwrite(output_path, result)


def main(args=None):
    if args is None:
        args = sys.argv[1:]

    parser = argparse.ArgumentParser(
        description='This script extracts all the contuors from images in a dir. It will crop the contours so they will all have 3 extra pixels around them. You can use it to extract glyphs from a document.')
    parser.add_argument('Images Folder', nargs='?',
                        help='Path to the folder that contains the images.', default=os.getcwd())
    args = parser.parse_args(args)
    # access positional argument using this syntax
    input_dir = vars(args)['Images Folder']

    for fileName in os.listdir(input_dir):
        if fileName.endswith((".jpg", ".jpeg")):
            extractAndSaveContoursFromImage(os.path.abspath(
                os.path.join(input_dir, fileName)), 10, 10, 100)


if __name__ == "__main__":
    import sys
    sys.exit(main())

The research for contextual spacing fonts was partially funded by the stimuleringsfonds.