Implementing various types of OpenCV functions with Python for Computer Vision tasks.
Theory and experimental results (on this page):
- The easiest ways to install OpenCV for Linux Ubuntu
- Brightness and Rotation changes
- Contrast Limited Adaptive Histogram Equalization
There are few the simplest ways to install OpenCV for Linux Ubuntu, and they are:
- by using
conda - by using
pip3for python 3
With first one it's needed to have conda been installed and to use one of the following commands:
conda install -c conda-forge opencvconda install -c conda-forge/label/broken opencv
With second one use one of the following commands:
pip3 install opencv-python- only main modulespip install opencv-contrib-python- both main and contrib modules
Also, might need to be installed following, if there is mistake arise:
sudo apt-get install libsm6sudo apt-get install -y libxrender-dev
OpenCV will be installed in choosen environment.
To check if OpenCV was installed, run python in any form and run following two lines of code:
import cv2
print(cv2.__version__)As a result, the version of installed OpenCV has to be shown, like this:
4.1.0
Easy way to rotate image and change brightness with OpenCV.
Can be used for preprocessing datasets for Classification Tasks.
# Importing needed libraries
import numpy as np
from PIL import Image
import matplotlib.pyplot as plt
import cv2
# Reading image with PIL library
image_PIL = Image.open('images/stop_sign.jpg')
# Converting read image into numpy array
image_PIL_np = np.array(image_PIL)
# Reading image with OpenCV library
# With OpenCV image is opened already as numpy array
# WARNING! OpenCV by default reads images in BGR format
image_OpenCV = cv2.imread('images/stop_sign.jpg', 1)
# Converting BGR to RGB
image_OpenCV = cv2.cvtColor(image_OpenCV, cv2.COLOR_BGR2RGB)
# Checking if arrays are equal
print(np.array_equal(image_PIL_np, image_OpenCV)) # True
# Showing image in OpenCV window
cv2.namedWindow('Traffic Sign', cv2.WINDOW_NORMAL) # Specifing that window is resizable
# Showing image
# WARNING! 'cv2.imshow' takes images in BGR format
# Consequently, we need to convert image from RGB to BGR
cv2.imshow('Traffic Sign', cv2.cvtColor(image_OpenCV, cv2.COLOR_BGR2RGB))
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('Traffic Sign')
# cv2.destroyAllWindows()
# Defining function for changing brightness
def brightness_changement(image):
# Converting firstly image from RGB to HSV
image_hsv = cv2.cvtColor(image, cv2.COLOR_RGB2HSV)
# Defining random value for changing brightness
random_brightness = np.random.uniform()
# Implementing changing of Value channel of HSV image
image_hsv[:, :, 2] = image_hsv[: , :, 2] * random_brightness
# Converting HSV changed image to RGB
image_rgb = cv2.cvtColor(image_hsv, cv2.COLOR_HSV2RGB)
# Returning image with changed brightness
return image_rgb
"""
To rotate an image using OpenCV Python,
first, calculate the affine matrix that does the affine transformation (linear mapping of pixels),
then warp the input image with the affine matrix.
Example:
M = cv2.getRotationMatrix2D(center, angle, scale)
rotated = cv2.warpAffine(img, M, (w, h))
where
center: center of the image (the point about which rotation has to happen)
angle: angle by which image has to be rotated in the anti-clockwise direction
scale: 1.0 mean, the shape is preserved. Other value scales the image by the value provided
rotated: ndarray that holds the rotated image data
"""
# Defining function for changing rotation of image
def rotation_changement(image, angle):
# Getting shape of image
rows, columns, channels = image.shape
# Implementing rotation
# Calculating Affine Matrix
affine_matrix = cv2.getRotationMatrix2D((columns / 2, rows / 2), angle, 1)
# Warping original image with Affine Matrix
rotated_image = cv2.warpAffine(image, affine_matrix, (columns, rows))
# Returning rotated image
return rotated_image
# Plotting RGB example with changes
# Pay attention!
# In OpenCV conventional ranges for R, G, and B channel values are:
# 0 to 255 for CV_8U images
# 0 to 65535 for CV_16U images
# 0 to 1 for CV_32F images
# Therefore if input image has values in [0..255] range,
# It is needed to convert it to uint8 dtype explicitly:
# im = np.random.randint(0,255,(224,224,3)).astype(np.uint8)
figure, ax = plt.subplots(nrows=2, ncols=4)
# 'ax 'is as (2, 4) np.array and we can call each time ax[0, 0]
# Plotting image with rotations
for i in range(4):
ax[0, i].set_axis_off()
ax[0, i].set_title(str(90 * (i +1)))
ax[0, i].imshow(rotation_changement(image_OpenCV, 90 * (i + 1)))
# Plotting image with changing brightness
for i in range(4):
ax[1, i].set_axis_off()
ax[1, i].imshow(brightness_changement(image_OpenCV))
plt.show()Results can be seen on the figure below.
Enhancing image by CLAHE algorithm with OpenCV.
Can be used for preprocessing datasets for Classification Tasks.
# Importing needed libraries
import numpy as np
import matplotlib.pyplot as plt
import cv2
import pickle
# Opening file for reading in binary mode
# Can be used any user's RGB image
with open('data0.pickle', 'rb') as f:
x_input = pickle.load(f, encoding='latin1') # dictionary type
# Getting image
x_rgb = x_input[4]
print(x_rgb.shape) # (32, 32, 3)
# Showing RGB image in OpenCV window
cv2.namedWindow('RGB', cv2.WINDOW_NORMAL) # Specifing that window is resizable
# WARNING! 'cv2.imshow' takes images in BGR format
# Consequently, we need to convert image from RGB to BGR
cv2.imshow('RGB', cv2.cvtColor(x_rgb, cv2.COLOR_RGB2BGR))
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('RGB')
# Converting image from RGB to LAB colour model
x_lab = cv2.cvtColor(x_rgb, cv2.COLOR_RGB2LAB)
# Showing LAB image in OpenCV window
cv2.namedWindow('LAB', cv2.WINDOW_NORMAL) # Specifing that window is resizable
cv2.imshow('LAB', x_lab)
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('LAB')
# Splitting LAB image to different channels
l, a, b = cv2.split(x_lab)
# Showing L channel in OpenCV window
cv2.namedWindow('L', cv2.WINDOW_NORMAL) # Specifing that window is resizable
cv2.imshow('L', l)
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('L')
# Showing A channel in OpenCV window
cv2.namedWindow('A', cv2.WINDOW_NORMAL) # Specifing that window is resizable
cv2.imshow('A', a)
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('A')
# Showing B channel in OpenCV window
cv2.namedWindow('B', cv2.WINDOW_NORMAL) # Specifing that window is resizable
cv2.imshow('B', b)
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('B')
# Applying CLAHE to L channel
clahe = cv2.createCLAHE(clipLimit=3.0, tileGridSize=(8,8))
l_clahe = clahe.apply(l)
# Showing L channel after CLAHE in OpenCV window
cv2.namedWindow('L CLAHE', cv2.WINDOW_NORMAL) # Specifing that window is resizable
cv2.imshow('L CLAHE', l_clahe)
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('L CLAHE')
# Merging enhanced L channel with CLAHE with other channels A and B
x_lab_enhanced = cv2.merge((l_clahe, a, b))
# Showing enhanced LAB image after CLAHE in OpenCV window
cv2.namedWindow('LAB Enhanced', cv2.WINDOW_NORMAL) # Specifing that window is resizable
cv2.imshow('LAB Enhanced', x_lab_enhanced)
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('LAB Enhanced')
# Converting image from LAB to RGB colour model
x_rgb_enhanced = cv2.cvtColor(x_lab_enhanced, cv2.COLOR_LAB2RGB)
# Showing enhanced RGB image in OpenCV window
cv2.namedWindow('RGB Enhanced', cv2.WINDOW_NORMAL) # Specifing that window is resizable
# WARNING! 'cv2.imshow' takes images in BGR format
# Consequently, we need to convert image from RGB to BGR
cv2.imshow('RGB Enhanced', cv2.cvtColor(x_rgb_enhanced, cv2.COLOR_RGB2BGR))
cv2.waitKey(0) # Waiting for any key being pressed
cv2.destroyWindow('RGB Enhanced')
cv2.destroyAllWindows()
%matplotlib inline
# Showing results with matplotlib
plt.rcParams['figure.figsize'] = (10.0, 8.0) # Setting default size of plots
figure, ax = plt.subplots(nrows=1, ncols=4)
# 'ax 'is as (1, 1) np.array and we can call each time ax[0, 0]
ax[0].set_axis_off()
ax[0].set_title('RGB')
ax[0].imshow(x_rgb)
ax[1].set_axis_off()
ax[1].set_title('LAB')
ax[1].imshow(x_lab)
ax[2].set_axis_off()
ax[2].set_title('LAB Enhanced')
ax[2].imshow(x_lab_enhanced)
ax[3].set_axis_off()
ax[3].set_title('RGB Enhanced')
ax[3].imshow(x_rgb_enhanced)
plt.show()Results can be seen on the figure below.
Valentyn N Sichkar. Neural Networks for computer vision in autonomous vehicles and robotics // GitHub platform. DOI: 10.5281/zenodo.1317904