Abstract
This work focuses on a parameter-free joint piecewise smooth image denoising and contour detection. Formulated as the minimization of a discrete Mumford–Shah functional and estimated via a theoretically grounded alternating minimization scheme, the bottleneck of such a variational approach lies in the need to fine-tune their hyperparameters, while not having access to ground truth data. To that aim, a Stein-like strategy providing optimal hyperparameters is designed, based on the minimization of an unbiased estimate of the quadratic risk. Efficient and automated minimization of the estimate of the risk crucially relies on an unbiased estimate of the gradient of the risk with respect to hyperparameters. Its practical implementation is performed using a forward differentiation of the alternating scheme minimizing the Mumford–Shah functional, requiring exact differentiation of the proximity operators involved. Intensive numerical experiments are performed on synthetic images with different geometry and noise levels, assessing the accuracy and the robustness of the proposed procedure. The resulting parameter-free piecewise-smooth estimation and contour detection procedure, not requiring prior image processing expertise nor annotated data, can then be applied to real-world images.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Data availability and materials
A MATLAB toolbox implementing the proposed automated image denoising and contour detection procedure is publicly available at https://github.com/charlesglucas/sugar_dms.
References
Bertasius, G., Shi, J., Torresani, L.: Deepedge: A multi-scale bifurcated deep network for top-down contour detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 4380–4389 (2015)
Cai, X., Steidl, G.: Multiclass segmentation by iterated ROF thresholding. In: International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, Springer, pp 237–250 (2013)
Deledalle, C.A., Vaiter, S., Fadili, J., et al.: Stein Unbiased GrAdient estimator of the Risk (SUGAR) for multiple parameter selection. SIAM J. Imag. Sci. 7(4), 2448–2487 (2014)
Foare, M., Pustelnik, N., Condat, L.: Semi-linearized proximal alternating minimization for a discrete Mumford–Shah model. IEEE Trans. Image Process. 29(1), 2176–2189 (2019)
Geman, S., Geman, D.: Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images. IEEE Trans. Pattern Anal. Match. Int. 6, 721–741 (1984)
Jaccard, P.: Distribution de la flore alpine dans le bassin des Dranses et dans quelques régions voisines. Bull. Soc. Vaudoise Sci. Nat. 37, 241–272 (1901)
Kiefer, L., Storath, M., Weinmann, A.: PALMS image partitioning-a new parallel algorithm for the piecewise affine-linear Mumford–Shah model. Image Process. On Line 10, 124–149 (2020)
Lucas, C., Pascal, B., Pustelnik, N, et al.: Hyperparameter selection for Discrete Mumford-Shah – Supplementary material (2022)
Mumford, D.B., Shah, J.: Optimal approximations by piecewise smooth functions and associated variational problems. Commun. Pure Appl. Math. 42(5), 577–685 (1989)
Pascal, B., Pustelnik, N., Abry, P.: Strongly convex optimization for joint fractal feature estimation and texture segmentation. Appl. Comp. Harm. Anal. 54, 303–322 (2021)
Pascal, B., Vaiter, S., Pustelnik, N., et al.: Automated data-driven selection of the hyperparameters for Total-Variation based texture segmentation. J Math Imaging Vis pp 1–30 (2021)
Ramani, S., Blu, T., Unser, M.: Monte–Carlo SURE: a black-box optimization of regularization parameters for general denoising algorithms. IEEE Trans. Image Process. 17(9), 1540–1554 (2008)
Roth, S., Black, M.J.: Fields of experts. Int. J. Comput. Vis. 82(2), 205–229 (2009)
Stein, C.: Estimation of the mean of a multivariate normal distribution. Ann. Stat. 9(6), 1135–1151 (1981)
Storath, M., Weinmann, A., Demaret, L.: Jump-sparse and sparse recovery using Potts functionals. IEEE Trans. Signal Process. 62(14), 3654–3666 (2014)
Zach, C., Häne, C.: Discretized convex relaxations for the piecewise smooth Mumford-Shah model. In: Proc. Energy Minimization Methods in Computer Vision and Pattern Recognition, pp 548–563 (2017)
Funding
This work is supported by ANR-19-CE48-0009s MULTISC-IN.
Author information
Authors and Affiliations
Contributions
The methodological content and contribution were developed equally by all the authors. The codes were mainly developed by C.G. Lucas with the help of B. Pascal and N. Pustelnik. The text was equally written and reviewed by all the authors. The figures and tables were mainly prepared by C-G. Lucas, with the help of N. Pustelnik.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no conflict of interest.
Ethical approval
Not applicable.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Lucas, CG., Pascal, B., Pustelnik, N. et al. Hyperparameter selection for Discrete Mumford–Shah. SIViP 17, 1897–1904 (2023). https://doi.org/10.1007/s11760-022-02401-1
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11760-022-02401-1