windows_cleaner.py

import cv2
import numpy as np
import pandas as pd


def delete_little_part(img,text_pixel_ratio:int=10,symbol_pixel_ratio:int=60,x_step:int=10,y_step:int=10,is_background_white:bool=True, crop_around:bool=True):
    #The explanation of the project's application is that the image is divided into separate and small areas with steps (x_step, y_step). Now, if there is text in any window, the number of black pixels compared to total pixels (assuming that the background is white) must be greater than a value (which we have specified as an input called "text_pixel_ratio") so that window should be Recognized as a window with text and it will not be cleaned.
    #In addition, the next idea is that if the number of black pixels on the white background of the image is greater than a limit (which we specified with the name "symbol_pixel_ratio") in this case, that window will be identified as the window containing the symbol and will be deleted.
    #It should be mentioned that the third idea implemented in this coding is that the input parameter called "crop_around" is defined, and when it is active, the final step in the windows of the x and y axes will always be cleared.
    x_step:int=max(20,x_step)
    y_step:int=max(30,y_step)
    if img.shape[2]>1:
        img = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
    cleaned_img=img.copy()
    #cleaned_img=cv2.threshold(cleaned_img, 50, 255, cv2.THRESH_BINARY )
    for y in range(0,img.shape[0],y_step):
        for x in range(0,img.shape[1],x_step):
            black_count:int=0
            white_count:int=0
            if (x==0 or y==0 or (x in range(0,img.shape[1],x_step)[-1:]) or (y in range(0,img.shape[0],y_step)[-1:])) and (crop_around):
                if is_background_white:
                    cleaned_img[y:y+y_step,x:x+x_step]=255
                else:
                    cleaned_img[y:y+y_step,x:x+x_step]=0
            else:
                for j in range(y,y+y_step):
                    for i in range(x,x+x_step):
                        if j<img.shape[0] and i<img.shape[1]:
                            if img[j,i]>=250:
                                white_count +=1
                            elif img[j,i]<=10:
                                black_count +=1
                total_pixel=x_step*y_step
                b_w_ratio:float=black_count/total_pixel
                w_b_ratio:float=white_count/total_pixel
                if is_background_white:
                    if b_w_ratio<text_pixel_ratio/100 or b_w_ratio>symbol_pixel_ratio/100:
                        cleaned_img[y:y+y_step,x:x+x_step]=255
                elif ~is_background_white:
                    if w_b_ratio<text_pixel_ratio/100 or w_b_ratio>symbol_pixel_ratio/100:
                        cleaned_img[y:y+y_step,x:x+x_step]=0
    return cleaned_img