mstorath / pottslab

Unsupervised multilabel image segmentation (color/gray/multichannel) based on the Potts model (aka piecewise constant Mumford-Shah model)

Pottslab

Pottslab is a Matlab/Java toolbox for the reconstruction of jump-sparse signals and images using the Potts model (also known as "piecewise constant Mumford-Shah model" or "l0 gradient model"). Applications include denoising of piecewise constant signals, step detection and segmentation of multichannel image.

-- See also the Pick of the Week on --

Application examples

Segmentation of vector-valued images

• Supports segmentation of vector-valued images (e.g. multispectral images, feature images)
• Linear complexity in number of color channels
• Label-free: No label discretization required

Left: A natural image; Right: Result using Potts model

Texture segmentation using highdimensional curvelet-based feature vectors

Used as segmentation method in

• A. Breger et al., Supervised learning and dimension reduction techniques for quantification of retinal fluid in optical coherence tomography images, Eye (2017).

Joint image reconstruction and segmentation

• Applicable to many imaging operators, e.g. convolution, Radon transform, MRI, PET, MPI: only implementation of proximal mapping reuqired - Supports vector-valued data - Label-free: Labels need NOT be chosen a-priori

Left: Shepp-Logan phantom; Center: Filtered backprojection from 7 angular projections; Right: Joint reconstruction and segmentation using the Potts model from 7 angular projections

Denoising of jump-sparse/piecewise-constant signals, or step detection/changepoint detection

• L1 Potts model is robust to noise and to moderately blurred data
• Fast and exact solver for L1 Potts model
• Approximative strategies for severely blurred data

Top: Noisy signal; Bottom: Minimizer of Potts functional (ground truth in red)

Used as step detection algorithm in

• A. Nord et al., Catch bond drives stator mechanosensitivity in the bacterial flagellar motor, Proceedings of the National Academy of Sciences, 2017
• A. Szorkovszky et al., Assortative interactions revealed by sorting of animal groups, Animal Behaviour, 2018

Usage Instructions

Standalone usage from command line (only image plain image segmentation supported)

• Call "java -jar pottslab-standalone.jar input output.png gamma" where gamma is a positive real number, e.g. 0.1 (thanks to fxtentacle)

Installation for Matlab (all features usable)

Quickstart:

• Run the script "installPottslab.m", it should set all necessary paths
• For best performance, increase Java heap space in the Matlab preferences (MATLAB - General - Java heap memory)
• Run a demo from the Demos folder

Troubleshooting:

• Problem: OutOfMemoryException

• Solution: Increase Java heap space in the Matlab preferences (MATLAB - General - Java heap memory)

• Problem: Undefined variable "pottslab" or class "pottslab.JavaTools.minL2Potts"

• Solution:

  • Run setPLJavaPath.m
  • Maybe you need to install Java 1.7 (see e.g. http://undocumentedmatlab.com/blog/using-java-7-in-matlab-r2013a-and-earlier)

Plugins for Image Analysis GUIs

Parts of Pottslab can be used without Matlab as pure Java plugins

• Icy plugin - an interactive image segmentation plugin based on Pottslab (written by Vasileios Angelopoulos)
• ImageJ plugin - an ImageJ frontend for Pottslab (written by Michael Kaul)

References

• M. Storath, A. Weinmann, J. Frikel, M. Unser. "Joint image reconstruction and segmentation using the Potts model" Inverse Problems, 2015
• A. Weinmann, M. Storath. "Iterative Potts and Blake-Zisserman minimization for the recovery of functions with discontinuities from indirect measurements." Proceedings of The Royal Society A, 471(2176), 2015
• A. Weinmann, M. Storath, L. Demaret. "The L1-Potts functional for robust jump-sparse reconstruction" SIAM Journal on Numerical Analysis, 2015
• M. Storath, A. Weinmann. "Fast partitioning of vector-valued images" SIAM Journal on Imaging Sciences, 2014
• M. Storath, A. Weinmann, L. Demaret. "Jump-sparse and sparse recovery using Potts functionals" IEEE Transactions on Signal Processing, 2014

See also

• PALMS Image Segmentation for piecewise affine linear estimation
• Higher order Mumford-Shah models for piecewise smooth signals
• Smoothing splines for discontinuous signals
• Mumford-Shah regularization for piecewise smooth images
• L1 TV for robust piecewise constant signals

pottslab's Issues

Getting a mask with segment labels

Is there a way to get an integer mask indicating the unique segment each pixel belongs to? That is, two pixels may end up with the same color in the Potts segmentation but, if they do not belong to the same contiguous region, should receive different labels. This is the result returned by the scikit image segmentation algorithms, for example.