PALMS_ImagePartitioning is a Matlab/C++ toolbox for image partitioning using the piecewise affine-linear Mumford-Shah model (also known as "affine-linear Potts model").
- Supports segmentation of vector-valued images (e.g. RGB images, multispectral images, feature images)
- Linear complexity in number of channels
- Label-free: no label discretization required
Left: natural image; Right: partitioning using the piecewise affine-linear Mumford-Shah model
- Avoids oversegmentation of images with linear trends (e.g. the sky in a landscape image, illumination gradients)
Left: natural image; Center: classical (piecewise constant) Potts model; Right: piecewise affine-linear Mumford-Shah model
- Scale of the partitioning is controlled by a model parameter
Large values of the model parameter give few segments, while small choices lead to more segments and closeness to the data
An online demo can be found at http://www.ipol.im/pub/art/2020/295/
The algorithm depends on a mex script that needs to be compiled before execution. For compilation inside MATLAB, cd into the 'src/cpp' folder and run build.m
Requires the Armadillo and OpenMP library
Tested with Armadillo 8.400 https://launchpad.net/ubuntu/+source/armadillo/1:8.400.0+dfsg-2 and OpenMP 4.0. On Linux, just use your package manager to install it:
sudo apt-get install libarmadillo-dev
sudo apt-get install libomp-dev
For a test run on the test image "redMacaw", run demo.m. demo.m calls the main function, affineLinearPartitioning.m Arguments of affineLinearPartitioning.m are:
- f: input image (double)
- gamma: boundary penalty (larger choice -> less segments)
- varargin: optional input parameters
- L. Kiefer, M. Storath, A. Weinmann. "An efficient algorithm for the piecewise affine-linear Mumford-Shah model based on a Taylor jet splitting." IEEE Transactions on Image Processing, 2020.
- L. Kiefer, M. Storath, A. Weinmann. "PALMS image partitioning – a new parallel algorithm for the piecewise affine-linear Mumford-Shah model." Image Processing On Line, 2020.