Plug and Play Algorithms for Imaging
(2024) In Bachelor’s Theses in Mathematical Sciences MATK11 20232Mathematics (Faculty of Engineering)
Mathematics (Faculty of Sciences)
Centre for Mathematical Sciences
- Abstract
- In this thesis, we explore the Alternating Direction Method of Multipliers (ADMM) algorithm, the Plug & Play (PnP) algorithm and consensus equilbrium in order to understand their importance for solving complex imaging problems. We begin by motivating the invention of ADMM which combines the strengths of dual ascent and the method of multipliers. A proof of convergence for ADMM is then provided. Afterwards we go on to show how ADMM can be used to solve the global consensus optimisation problem and how PnP builds upon ADMM and extends its capabilities. However, using PnP means we no longer have guaranteed convergence. To remedy this the Consensus Equilibrium (CE) framework is introduced, which is an optimisation free framework that... (More)
- In this thesis, we explore the Alternating Direction Method of Multipliers (ADMM) algorithm, the Plug & Play (PnP) algorithm and consensus equilbrium in order to understand their importance for solving complex imaging problems. We begin by motivating the invention of ADMM which combines the strengths of dual ascent and the method of multipliers. A proof of convergence for ADMM is then provided. Afterwards we go on to show how ADMM can be used to solve the global consensus optimisation problem and how PnP builds upon ADMM and extends its capabilities. However, using PnP means we no longer have guaranteed convergence. To remedy this the Consensus Equilibrium (CE) framework is introduced, which is an optimisation free framework that generalises the global consensus optimisation problem. The thesis concludes by showing ADMM as a special case of the CE equations. (Less)
- Popular Abstract (Swedish)
- I den digitala tidsåldern spelar invecklad matematik ofta en undanskymd, men ändå avgörande roll. När vi löser bildproblem, som att avpixla foton och förbättra upplösningen, använder vi ofta komplex matematik. I den här avhandlingen vill vi lyfta fram den genom att beskriva tre viktiga algoritmer - ADMM, Plug and Play och Consensus Equilibrium. Denna avhandling utforskar hur dessa algoritmer bygger på varandra, deras sammankopplingar och deras underliggande betydelse.
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9144589
- author
- Hughes, Alexander LU
- supervisor
- organization
- alternative title
- Plug and Play-algoritmer för bildbehandling
- course
- MATK11 20232
- year
- 2024
- type
- M2 - Bachelor Degree
- subject
- keywords
- ADMM, Plug & Play algorithm, consensus equilibrium, consensus optimisation
- publication/series
- Bachelor’s Theses in Mathematical Sciences
- report number
- LUNFMA-4154-2024
- ISSN
- 1654-6229
- other publication id
- 2024:K1
- language
- English
- id
- 9144589
- date added to LUP
- 2025-06-27 15:57:06
- date last changed
- 2025-06-27 15:57:06
@misc{9144589, abstract = {{In this thesis, we explore the Alternating Direction Method of Multipliers (ADMM) algorithm, the Plug & Play (PnP) algorithm and consensus equilbrium in order to understand their importance for solving complex imaging problems. We begin by motivating the invention of ADMM which combines the strengths of dual ascent and the method of multipliers. A proof of convergence for ADMM is then provided. Afterwards we go on to show how ADMM can be used to solve the global consensus optimisation problem and how PnP builds upon ADMM and extends its capabilities. However, using PnP means we no longer have guaranteed convergence. To remedy this the Consensus Equilibrium (CE) framework is introduced, which is an optimisation free framework that generalises the global consensus optimisation problem. The thesis concludes by showing ADMM as a special case of the CE equations.}}, author = {{Hughes, Alexander}}, issn = {{1654-6229}}, language = {{eng}}, note = {{Student Paper}}, series = {{Bachelor’s Theses in Mathematical Sciences}}, title = {{Plug and Play Algorithms for Imaging}}, year = {{2024}}, }