This code is an implementation of Gaussian and Laplacian pyramids and the fusion visible and NIR images according to Vanmali et al article.

This code works with Python 3 and requires the following libraries:

• matplotlib pip install matplotlib
• skimage pip install skimage
• numpy pip install numpy
• sklearn pip install sklearn
• scipy pip install scipy

To get the Gaussian and Laplacian pyramids of an image as well as the reconstruction of an image, run the following script:

python main.py

You may want to change the path of the image in the script and call the function main_gaussian_laplacian_pyramids(image, kernel, levels).

The obtained results are below for the Lenna example:

Original image

Gaussian pyramid

Laplacian pyramid

Reconstruction

This program works for grayscale and three channels images.

All the equations used to implement NIR and visible fusion can be found in Vanmali et al article.

To achieved image visibility enhancement using fused Laplacian pyramids applied to NIR and visible images, run the following script:

python main.py

You may want to change the path of the image in the script and call the function main_multimodal_fusion(im_vis, im_ir, kernel, levels, window_size).

The obtained results are below for the landscape example:

On the left, value channel from HSV, on the right NIR image

From left to right : local entropy, local contrast and visibility. First row is value channel from HSV and second row is NIR image

Weights normalization, on the left value channel weight from HSV and on the right NIR image weight

Fused Laplacian pyramid

On th left, visible RGB image before transformation, on the right, NIR and visible image fusion displayed in the visible RGB format