Image normalization is a preprocessing technique that is commonly used in computer vision applications. The goal of image normalization is to transform the pixel values of an image that are within a certain range or have certain statistical properties. Normalization is used to reduce the impact of variations in lighting conditions or to standardize the color or brightness of images.
Normalization techniques typically involve scaling the pixel values of an image to fall within a certain range or modifying the distribution of pixel values to have certain statistical properties. There are many different techniques for image normalization, and the choice of technique depends on the specific application and the characteristics of the image data.
Here are some common techniques for image normalization.
Min-max normalization: This technique scales the pixel values of an image so that they fall within a specified range, typically [0, 1] or [-1, 1]. This can be done using the following formula:
normalized_image = (image – min_value) / (max_value – min_value)
Here, min_value and max_value are the minimum and maximum pixel values in the image, respectively.
Z-score normalization: This technique modifies the distribution of pixel values in an image to have a mean of 0 and a standard deviation of 1. This can be done using the following formula:
normalized_image = (image – mean_value) / std_value
Here, mean_value and std_value are the mean and standard deviation of the pixel values in the image, respectively.
Histogram equalization: This technique modifies the distribution of pixel values in an image to be more uniform. This can be done by computing the cumulative distribution function (CDF) of the pixel values and mapping the pixel values to new values based on the CDF:
import cv2
# Load image
img = cv2.imread(“image.jpg”, 0)
# Apply histogram equalization
equalized_img = cv2.equalizeHist(img)
In the preceding code, we first load an image using the OpenCV library. We then apply histogram equalization using the equalizeHist() function, which returns a new image with a more uniform distribution of pixel values. OpenCV is a powerful and widely used open source library that plays a crucial role in image recognition and computer vision tasks. Its importance stems from its comprehensive collection of tools, functions, and algorithms designed to handle various aspects of image processing, analysis, and recognition.
Let’s see an example of image normalization using Python. We first import the necessary libraries: os for file and directory operations, cv2 for image loading and manipulation, and numpy for mathematical operations:
import os
import cv2
import numpy as np
We define the path to the image directory and get a list of all image filenames in the directory using a list comprehension:
# Define the path to the image directory
img_dir = ‘path/to/image/directory’
# Get a list of all image filenames in the directory
img_files = [os.path.join(img_dir, f) \
for f in os.listdir(img_dir) \
if os.path.isfile(os.path.join(img_dir, f))]
We loop through all the image files using a for loop. For each image file, we load the image using OpenCV (cv2.imread()):
# Loop through all the image files
for img_file in img_files:
# Load the image using OpenCV
img = cv2.imread(img_file)
We convert the image to float32 data type using astype(np.float32). This is necessary for the next step of normalization:
# Convert the image to float32 data type
img = img.astype(np.float32)
We normalize the image pixels to have zero mean and unit variance using the following formula: img -= np.mean(img); img /= np.std(img). This is also known as standardization or z-score normalization. This step is important for machine learning models that are sensitive to the scale of input features, as it ensures that the pixel values have a similar scale across all images:
# Normalize the image pixels to have zero mean and unit variance
img -= np.mean(img)
img /= np.std(img)
Finally, we save the normalized image with the same filename using cv2.imwrite():
# Save the normalized image with the same filename
cv2.imwrite(img_file, img)
Image normalization is a critical step in many computer vision applications, as it can help to reduce the impact of variations in lighting conditions and standardize the color and brightness of images. By transforming the pixel values of an image, we can make it easier for machine learning algorithms to learn from the image data and improve the accuracy of our models.