motion_blur.py

# Import the dependencies
import argparse
import cv2 #image processing lib
import numpy as np #matrix manipulation
from time import sleep #for slowing down the process to make progress visible 
from tqdm import tqdm as tqdm # for progess bar 

def motionBlur(img):
    # default kernel size.
    size = 70
    # Size of kernel is directly proportional to effect of blur
    # Create the Horizontal and Vertical kernel. 
    Hkernel = np.zeros((size, size))
    Vkernel = np.copy(Hkernel)

    # Fill the middle row with ones.  
    Hkernel[int((size - 1)/2), :] = np.ones(size) 
    Vkernel[:, int((size - 1)/2)] = np.ones(size)
        
    # Normalize the value under range of 0-1 
    Hkernel /= size 
    Vkernel /= size
    # Apply the vertical kernel. 
    v = cv2.filter2D(img, -1, Vkernel) 

    # Apply the horizontal kernel. 
    h = cv2.filter2D(img, -1, Hkernel)
    # Create Diagonal filter kernel
    Dkernel = np.copy(Hkernel)
    i=0 #initialize index matching var.
    for x in tqdm(Dkernel):
        for j in range(0,size):
            if(j==i):
                x[i]=1
        i+=1
        sleep(0.01) # to make the progress visible
    # Normalize. 
    Dkernel /= size
    
    # Apply the Diagonal kernel. 
    d = cv2.filter2D(img, -1, Dkernel)
    
    return v,h,d
    
# construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-i", "--image", required=True, help="path to input image")
args = vars(ap.parse_args())

img = Image.open(args['image'])

print("Work in Progress.")
# Save the outputs. 
v,h,d=motionBlur(img)
cv2.imwrite('assets/horizontalBlur.jpg', v) 
cv2.imwrite('assets/verticalBlur.jpg', h)
cv2.imwrite('assets/diagonalBlur.jpg', d) 
print("Work done successfully!")