crop lecture notes that is written on white paper

lecture_notes_croper.py

import argparse as arg
import os

import cv2
import numpy as np


def generate_cropped_by_white(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    blurred = cv2.GaussianBlur(gray, (35, 35), 0)
    _, thr = cv2.threshold(blurred, 200, 255,
                           cv2.THRESH_BINARY + cv2.THRESH_OTSU)
    contours, _ = cv2.findContours(thr.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)
    contour_areas = np.array(map(cv2.contourArea, contours))
    max_contour = np.argmax(contour_areas, axis=0)
    x, y, w, h = cv2.boundingRect(contours[max_contour])
    img = img[y:y + h, x:x + w]
    return img, img_path


if __name__ == '__main__':
    parser = arg.ArgumentParser(usage="python this.py --images image1.png ~/Downloads/test.png --out ./test_dir")
    parser.add_argument('image_paths', metavar='path', type=str, nargs='+', help='path to images separated with space')
    parser.add_argument('--out', type=str, help='path to outputdir')
    args = parser.parse_args()

    for (cropped, orig_path) in map(generate_cropped_by_white, args.image_paths):
        file_name = os.path.basename(orig_path)
        print "proccess image %s" % file_name
        cv2.imwrite(os.path.join(args.out, file_name + '.gen.jpg'), cropped)