xiaoshulittletree/invert-shadowgraphy

This is matlab codes to invert shadowgraphy or proton radiography images to obtain the deflection of the light beam (or proton beam in proton radiography case).

For the fast invert shadowgraphy, please see the fast-inverse branch

Invert shadowgraphy and proton radiography

This repository contains MATLAB codes to invert shadowgraphy images (or proton radiography images) to get its deflection potential of a diagnosed object. The description about this method and some assumptions can be found on http://arxiv.org/abs/1607.04179. It contains 3 main files:

• main_forward
• main_inverse_extended
• invert_shadowgraphy

In this code, I am using a normalised unit, where every pixel has size of 1 x 1, the distance from the object to the screen is 1, and the magnification is 1. The normalised deflection potential () and the deflection potential in normal units () are related by

where is the distance between the object and the screen, is the pixel size on the object plane, and is the magnification.

The codes contain other codes from:

• https://uk.mathworks.com/matlabcentral/fileexchange/56633-bounded-power-diagram
• https://uk.mathworks.com/matlabcentral/fileexchange/34428-voronoilimit
• https://uk.mathworks.com/matlabcentral/fileexchange/44385-power-diagrams
• https://www.cs.ubc.ca/~schmidtm/Software/minFunc.html

Getting started

Requirements

• MATLAB R2013b or later
• Parallel Computing toolbox (MATLAB)
• Mapping toolbox (MATLAB)
• Git

Steps

• Open cmd / bash / terminal and type the codes below

cd [the-directory-you-want-to-put-the-codes]
git clone https://github.com/mfkasim91/invert-shadowgraphy
cd invert-shadowgraphy
matlab

• (Optional) To run the demo, type demo in MATLAB
• To run the inversion in your computer, you can type:

invert_shadowgraphy('directory/to/your/proton/radiogram/image');

Submitting to clusters

Requirements

• MCR 8.3

Steps

• Copy the executable in the /bin folder to your cluster
• In the script to submit the job, include this lines:

module load matlab/mcr/8.3
./invert_shadowgraphy /complete/path/to/your/image <options>

Options

• fdir_out: the directory to put the results (default: pwd)
• num_sites: number of sites (default: min(100000, number of pixels))
• source_map:
  • 0 - uniform with the same size as the file (default: 0)
  • positive number - using tvdenoise of the input image with lambda = source_map argument
  • string - filename of the source
• algorithm: optimization algorithm, 'quasi-newton' or 'lbfgs' (default: 'lbfgs')
• num_workers: number of workers in the distributed computing (default: 12)