MultiScale Image Contrast Amplification (MUSICA)- Python Implementation

Contrast help us to differentiate different region of interest in radiology. An optimal contrast differential will aid physicians in diagnosis by improving the visibility of organs, vessels, bones, tissues etc, which in turn increase the probability of observing abnormality from normalcy.

Code is available here.

This code is a Python implementation of the classic paper titled Multiscale Image Contrast Amplification (MUSICA) by Pieter Vuylsteke & Emile Schoeters.

Firstly an image is decomposed into a multiscale representation. This can be Haar Transform, Wavelet Transform or, Laplacian Pyramid, the one which I have used in this implementation. Contrast Enhancement is achieved by amplifying this transformed coefficients non-linearly. Then these modified coefficients are reconstructed to obtain the enhanced image.

The non-linear amplification used is as follows:

Here x represents the original Laplacian coefficent.

After downloading the github repo we need to install the packages using pip:

$pip install -r requirements.txt
$python demo.py

You can play with parameters a, L and p given in the beginning of the demo script. Where a is amplification factor, L is the total number of levels in the Laplacian pyramid, M is the upper bound of the coefficients and p controls the degree of non-linearity.