image-dehazing

An implementation of the IEEE paper titled Effective Single Image Dehazing by Fusion by Codruta Orniana Ancuti, Cosmin Ancuti and Philippe Bekaert.

Contributors

For any queries, you may contact:

• Adhithya Arun ([email protected])
• Namratha Gopalabhatla ([email protected])

Directory Structure

• image-dehazing
• image_dehazing : Image dehazing package
  • __init__.py
  • single_image.py : ImageDehazing class
• demo.py : Demo script
• height_variance.py : Sample script to display variance in quality of dehazed results with height of image pyramid
• LICENSE : License
• README.md : Documentation
• requirements.txt : Package requirements

Pre-requisites

Below are the pre-requisites to be satisfied before the package can be used.

• Make sure python3 is installed. (Preferably python3.7)

• Package requirements:

  • matplotlib
  • numpy
  • opencv-python
  • scikit-image

• If the above packages are not installed,

  • You may manually install the packages individually
  • You can also install the packages using the requirements.txt:

    sudo python3.7 -m pip install -r requirements.txt

Class Structure

• ImageDehazing([verbose = False])
• Methods:

  • white_balance(image)
    Description: Function to perform white balancing operation on an image
    

    Parameter	Description	Default
    image	Image to be white balanced	None

  • enhance_contrast(image)
    Description: Function to enhance contrast in an image
    

    Parameter	Description	Default
    image	Image to be contrast enhanced	None

  • luminance_map(image)
    Description: Function to generate the Luminance Weight Map of an image
    

    Parameter	Description	Default
    image	Image whose luminance map is to be generated	None

  • chromatic_map(image)
    Description: Function to generate the Chromatic Weight Map of an image
    

    Parameter	Description	Default
    image	Image whose chromatic map is to be generated	None

  • saliency_map(image)
    Description: Function to generate the Saliency Weight Map of an image
    

    Parameter	Description	Default
    image	Image whose saliency map is to be generated	None

  • image_pyramid(image, pyramid_type, levels)
    Description: Function to generate the Gaussian/Laplacian pyramid of an image
    

    Parameter	Description	Default
    image	Image whose pyramid is to be generated	None
    pyramid_type	Type of pyramid: 'gaussian' or 'laplacian'	'gaussian'
    levels	Height of pyramid	1

  • fusion(inputs, weights, gaussians)
    Description: Function to fuse the pyramids together
    

    Parameter	Description	Default
    inputs	Images to be fused	None
    weights	Image to be white balanced	None
    gaussians	Gaussian pyramids for the corresponding inputs	None

  • dehaze(image, [verbose = None, pyramid_height = 12])
    Description: Driver function to dehaze the image
    

    Parameter	Description	Default
    image	Image to be dehazed	None
    verbose	Flag denoting whether each step should be displayed	None
    pyramid_height	Height of image pyramids	12

Usage

• Once the pre-requisites are satisfied, the demo can be run by running the following command:

  python3.7 demo.py

• Place the image_dehazing directory in the same directory as your code files that are using them.

• The ImageDehazing class from the package can be used as shown in the below example:

  from skimage.io import imread
  import matplotlib.pyplot as plt
  
  # Import ImageDehazing class
  from image_dehazing.single_image import ImageDehazing
  
  # Read image
  path_to_image = './dataset/image_1.jpg'  # Path to hazy image
  hazy_image = imread(path_to_image)
  
  # Create object of ImageDehazing class
  dehazer = ImageDehazing(verbose=True)
  
  # Dehaze the the image using th dehaze method of the object
  dehaze_data = dehazer.dehaze(hazy_image, pyramid_height=12)
  
  # Display dehazed image
  plt.figure()
  plt.subplot(1, 2, 1)
  plt.imshow(dehaze_data['hazed'])
  plt.title('Hazy Image')
  plt.axis('off')
  plt.subplot(1, 2, 2)
  plt.imshow(dehaze_data['dehazed'])
  plt.title('Dehazed Image')
  plt.axis('off')
  plt.show()

• To see results on analysis of the changes observed due to variation in height of the pyramid, run the height_variance.py script as follows:

  python3.7 height_variance.py