tringholm / dg-elastica

EulerElastica: Euler’s elastica-based image denoising and inpainting in MATLAB.

Recommended packages/tools:

• MATLAB Image Processing Toolbox for running the demo file.
• MATLAB Parallel Computing Toolbox for running code in parallel.
• MATLAB compatible C compiler for MEX functions. Parallel code will not work without MEX.

EulerElastica is a MATLAB based image analysis tool for denoising and/or inpainting damaged images using Euler’s elastica as a regularisation prior in a variational image analysis setting. The algorithm used for minimizing the resulting energy is the Discrete Gradient algorithm outlined in [1].

To get started, open the elasticaDemo.m file and read the commented code there.

The package contains 12 MATLAB files and 5 C files, with the following structure:

• eulerElastica.m is the main file, and the only one you need to use to apply the algorithm
• defaultOptions.m supplies an option struct which may be useful to tweak parameters
• elasticaDemo.m is a demonstration file showing how to use the algorithm with examples
• eulerElasticaMatlab/Mex/MexPara.m are lower level files executing the algorithm in MATLAB, MEX and Parallel versions, respectively
• coordFxn.m, dgstep.m, energyFxn.m, fzeroFast.m, gradCurv.m, and gradTV.m are all auxiliary functions used mostly in eulerElasticaMatlab.m
• dgstepMex/MexPara.c, gradCurvMex/MexPara.c and partitionMex.c are C functions for speeding up critical parts of the code in a MEX environment.

This is the first release, so if you have comments and suggestions for improvements, please contact me and I will try to accommodate them.

[1] V. Grimm, R. I. McLachlan, D. I. McLaren, G. Quispel, and C. Schönlieb, “Discrete gradient methods for solving variational image regular- isation models,” Journal of Physics A: Mathematical and Theoretical, vol. 50, no. 29, p. 295201, 2017.