Hyperspectral Approximation and Alignment Tool

This software allows the user to take a raw NEF file from a Nikon D5000 that has been dispersed through a prism. Using the spectral response of that camera's particular CMOS sense as a parameter, the dispersionImg class will align the content and then provide an approximation of the hyperspectral content of the scene without the need for a diffraction grating or hyperspectral sensor.

Usage

• Place the raw NEF image you want to undisperse into the /img directory.
• Set the software in main.py to look for that image
• Run the software using Python main.py

It will populate the image directory with an _output folder that is prefaced with the name of the input file. This directory will include 5 images:

• 3-channel-stacked-RGB: This is the image after having 3 of its channels aligned.
• 6-channel-multispectral-approximation: This is a set of 6 images that breaks down a few intermediate colors based on the aligned image and the provided color weights.
• A series of three images that provides a visualization of the matchpoints made. Each match is made from the green_1 channel to the other 3 channels (blue, green_2, and red).

You may need to adjust the parameters to be more or less permissive in main.py. Ideally you want a small number of accurate matches from green_1 to the other channels. (though green_1 to green_2 will typically produce many matches as they are very similar sensors)

Roadmap

This software is deeply in-progress. It was produced under a tight time constraint and has accumulated some technical debt I am working my way through. Though this project has had to be put on the backburner for the moment I will eventually be able to address the following concerns:

• Provide more "Pythonic" error handling to use exceptions rather than error flags.
• Provide the ability to apply to more than one type of RAW file taken by one type of camera (currently only .NEF from a Nikon D5000).
• Provide a graphical user interface.
• Separate the concerns as the main class dispersionImg is fairly large. If possible, I would like to refactor it to several classes.
• Provide functionality to look at the number of matches and try lower/higher parameter values to get better outputs.

Installation and Environment

Python 3.9.13 is necessary. Newer versions of Python will not work with the software.

The libraries below will also be required.

• numpy
• rawpy
• cv2 (opencv-python)
• matplotlib
• scipy

If you are working in a Windows environment Anaconda is suggested!