preprocessing_images.py

#importing packages
from pdf2image import convert_from_path
from fpdf import FPDF
import cv2
import numpy as np 
import requests
from PIL import Image
from PIL import ImageEnhance

def boxcutter(img):
    img = np.array(img)
    height, width = img.shape[:2]
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # convert to grayscale
    blurred = cv2.GaussianBlur(gray, (5, 5), 0)
    (T, threshInv) = cv2.threshold(blurred, 230, 255, cv2.THRESH_BINARY_INV)
    contours, hierarchy = cv2.findContours(threshInv,cv2.RETR_LIST, \
                                       cv2.CHAIN_APPROX_SIMPLE)
    mx = (0,0,0,0)      # biggest bounding box so far
    mx_area =0
    my_dict = {}
    for cont in contours:
        x,y,w,h = cv2.boundingRect(cont)
        area = w*h
        nw = x,y,w,h
        my_dict[nw] = area
    if len(my_dict)<=2:
        return img
    else:
        x,y,w,h = sorted(my_dict, key=my_dict.get)[-2]
        roi=img[y:y+h,x:x+w]
        height_r, width_r = roi.shape[:2]
        if height_r > int(height*0.6) and width_r > int(width*0.6):
            return roi
    #         cv2.imwrite('C:/Users/Meet/Desktop/crop/'+ img_path.split('/')[-1], roi)
    #         cv2.rectangle(img,(x,y),(x+w,y+h),(200,0,0),2)
    #         cv2.imwrite('C:/Users/Meet/Desktop/cont/' + img_path.split('/')[-1], img)

        else:
            return img
    #         cv2.imwrite('C:/Users/Meet/Desktop/crop/'+ img_path.split('/')[-1], img)
    #         cv2.rectangle(img,(x,y),(x+w,y+h),(200,0,0),2)
    #         cv2.imwrite('C:/Users/Meet/Desktop/cont/' + img_path.split('/')[-1], img)

def noise_removal(image):
    #increasing contrast of text
    image=Image.fromarray(image)
    enhancer=ImageEnhance.Contrast(image)
    enhanced_image=enhancer.enhance(2.5)
    image=np.array(enhanced_image)
    #noise removel using dilate,erode, morphology and median blur
    kernel = np.ones((1, 1), np.uint8)
    image=cv2.dilate(image,kernel,iterations=1)
    kernel=np.ones((1,1),np.uint8)
    image=cv2.erode(image,kernel,iterations=1)
    image = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel,iterations=1)
    image=cv2.medianBlur(image,3)
    image=cv2.bitwise_not(image)
    kernel=np.ones((1,1),np.uint8)
    image=cv2.erode(image,kernel,iterations=3)
    image=cv2.bitwise_not(image)
    image=cv2.bitwise_not(image)
    kernel=np.ones((1,1),np.uint8)
    image=cv2.dilate(image,kernel,iterations=3)
    image=cv2.bitwise_not(image)
    return (image)

def preprocessing_function(url):
    #reading and converting the pdf into image format
    response = requests.get(url)
    with open('metadata.pdf', 'wb') as f:
        f.write(response.content)
    images = convert_from_path('metadata.pdf', 200)
    #removing red color from the answer paper
    for i in range(len(images)):
        img = boxcutter(images[i])
        image = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
        img_hsv=cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
        lower_red = np.array([161,50,50])
        upper_red = np.array([189,255,255])
        mask1 = cv2.inRange(img_hsv, lower_red, upper_red)
        image[np.where(mask1==255)] = 255
        #using noise removal function to remove noise and enhance the images
        img_bw = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
        no_noise = noise_removal(img_bw)
        cv2.imwrite("no_noise"+str(i)+".jpg", no_noise)
    #saving the the final preprocessed images as pdf    
    pdf = FPDF()
    for j in range(len(images)):
        pdf.add_page()
        pdf.image("no_noise"+str(j)+".jpg", 0, 0, 200, 300)
    pdf.output("answer_paper.pdf", "F")