Image thresholding

No account is required. Learn the material, try the examples, and mark it complete when you are ready to move on. Sign in with email only if you want to save progress across devices.

Codepython

import cv2
import os
import matplotlib.pyplot as plt
import numpy as np

import urllib.request

from IPython.display import display, Image

Codepython

# If the folder "resources" exists, do nothing, else create the folder "resources"

resources_folder = "resources"

if os.path.isdir(resources_folder):
    print(f"Directory: `{resources_folder}` already exists")
else:
    print(f"Creating directory: `{resources_folder}`")
    os.makedirs(resources_folder)

Output

Directory: `resources` already exists

Codepython

# iruma_army_url = "https://github.com/cookieSensei/AI-tutorials/blob/main/iruma%20kun.PNG?raw=true"

# Creating CLI path to the image
iruma_army_img_path = os.path.join("resources", "iruma kun.PNG")
iruma_army_img_path

Codepython

def cv2_imshow(img_arr):
    """Function to visualize an image loaded using opencv.
    
    Parameters:
        img_arr: image array ing BGR format or say image loaded using opencv
    """
    
    img_arr = img_arr.copy()    
    image_rgb = cv2.cvtColor(img_arr, cv2.COLOR_BGR2RGB) 

    plt.figure(figsize=(6,6))
    plt.imshow(image_rgb)
    plt.show()

Codepython

sample_kernel = np.ones((9,9), np.float32)/81
sample_kernel

Output

array([[0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568],
       [0.01234568, 0.01234568, 0.01234568, 0.01234568, 0.01234568,
        0.01234568, 0.01234568, 0.01234568, 0.01234568]], dtype=float32)

Codepython

# Vertical blur
gaussian_blur4 = cv2.GaussianBlur(src=img, ksize=(15,15),
                                 sigmaX=0.1,
                                 sigmaY=20)
cv2_imshow(gaussian_blur4)

Output

Codepython

bilateral_filtering = cv2.bilateralFilter(src=img,
                                          d=20,
                                          sigmaColor=10,
                                          sigmaSpace=10)

bilateral_filtering_rgb = cv2.cvtColor(bilateral_filtering, cv2.COLOR_BGR2RGB)
plt.imshow(bilateral_filtering_rgb)

Output

<matplotlib.image.AxesImage at 0x7b6107a061d0>

Codepython

# Standardized or normlized image would mean that
# the maximum value of pixel brightness is 1 instead of 255
# We take the whole range of brightness from 0-255 and
# just normalize it into 0-1 by dividing each pixel numeric by 255
standardized_image_rgb[0, 0]

Output

array([1., 1., 1.])

Codepython

# Downsizing the image to 50% of its width and 40% of its height
img = cv2.resize(img_nlp_rgb, (int(width*0.5), int(height*0.4)),
                 interpolation=cv2.INTER_LINEAR)

img.shape

Output

(1612, 1134, 3)

Find your way

Image thresholding

Kernel and Blur

Gaussian blur

Vertical gaussian blur

Median Blur

Bilateral Filtering

Image array normalization

Thresholding

Capture pixel values from a given range

Very Dark Image threshold

Adaptive threshold

Bright Image Threshold

Lesson resources

Finished this lesson?