main.py

import cv2
import pytesseract
import numpy as np


# read the image to be processed
img = cv2.imread("test3.jpg")
#cv2.imshow("img", img)

#cv2.setMouseCallback("img", perspective, [img])
# apply perspective transformation to get a top view of the cards (from a 1920x1080 ss only)

matrix = cv2.getPerspectiveTransform(np.float32([
    [0, 1056/2], 
    [1056, 1056/2],     
    [0, 1056],
    [1056, 1056]]), 
    np.float32([[0,0], [1056,0], [0, 1056], [1056, 1056]]))
img = cv2.warpPerspective(img, matrix, (1056, 1056))

img_gray = img.copy()

# remove all the non-white pixels to identify the cards
height, width, channels = img.shape
for x in range(width):
    for y in range(height):
        if sum(img_gray[y,x]) < sum((170,170,170)):
            img_gray[y,x] = (0,0,0)

# find contours of white chunk of pixels
img_gray = cv2.cvtColor(img_gray, cv2.COLOR_BGR2GRAY)
ret, im = cv2.threshold(img_gray, 0, 255, cv2.THRESH_BINARY)
contours, hierarchie = cv2.findContours(img_gray, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)

# returns the original image but only the parts inside the selected contours
mask = np.zeros_like(img)
for contour in contours:
    if cv2.contourArea(contour) > 500:
        cv2.fillPoly(mask, [contour], (255,255,255))

result = cv2.bitwise_and(img, mask)

#cv2.imshow("img", result)
#cv2.imwrite("test3.jpg", result)
cv2.imwrite("test5.jpg", img)


cv2.waitKey(0)
cv2.destroyAllWindows()