Pottslab

Port Attribution

This Python/Rust package (pottslab) is an automated port of the original MATLAB/Java pottslab library:

Original authors	Martin Storath, Andreas Weinmann
Original license	MIT
Original language	MATLAB (high-level), Java (performance core)

Port details

Port author	Claude Sonnet coding agent (Anthropic, 2026)
Port language	Python (API layer) + Rust/PyO3 (performance core)
Port license	MIT (unchanged)

What was ported

All algorithms are faithful translations of the original MATLAB/Java implementations:

• src/l2potts.rs — Direct port of Java/src/pottslab/L2Potts.java::call(): O(n²) dynamic programming solver for the L2-Potts functional with early-termination acceleration and piecewise-constant reconstruction.
• src/l1potts.rs — Direct port of Java/src/pottslab/IndexedLinkedHistogram.java: sorted linked-list histogram for efficient L1-Potts DP.
• src/processor.rs — Direct port of Java/src/pottslab/PLProcessor.java: parallel (Rayon) application of 1D Potts in horizontal, vertical, diagonal, and anti-diagonal directions.
• src/admm.rs — Direct port of Java/src/pottslab/JavaTools.java::minL2PottsADMM4/8(): ADMM splitting for the 2D Potts problem with 4- and 8-connected neighborhoods.
• pottslab/inverse.py — Direct port of Potts/PottsCore/iPottsADMM.m: inverse Potts solver for γ‖Du‖₀ + ‖Au-f‖ₚ with scipy linear sub-solvers.
• pottslab/sparsity.py — Port of Sparsity/SparsityCore/minSpars.m.
• pottslab/tikhonov.py — Port of Tikhonov/minL2Tikhonov.m and minL1Tikhonov.m.
• pottslab/utils.py — Port of helpers in Auxiliary/.

What changed

• API names follow Python conventions (min_l2_potts instead of minL2Potts).
• Data is passed as NumPy arrays (float64) instead of MATLAB matrices.
• Performance-critical loops are in Rust (replacing the Java JAR).
• The 2D ADMM and directional processor use Rayon for parallelism.
• The Python API layer uses keyword-only arguments with the same defaults as MATLAB.

Academic references (from original authors)

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

This site is open source. Improve this page.