Guest Editors:
Tatiana Gelvez-Barrera: Université de Lyon, Lyon, France
Marcus Carlsson: Centre for Mathematical Sciences, Lund University
Carlos Hinojosa: King Abdullah University of Science and Technology (KAUST)
Sayantan Dutta: Weill Cornell Medicine, New York, New York, USA
Submission Status: Open | Submission Deadline: 29 February 2024
This special issue of the EURASIP Journal on Image and Video Processing invites leading researchers and practitioners in academia and industry to contribute original research on recent theoretical advances, algorithmic developments, and practical applications of the plug-and-play framework. Topics include new denoising operators, adaptive plug-and-play algorithms, convergence analysis, hyperparameter tuning, robustness, stability analysis, and signal, imaging, and video processing applications.
Model-based inverse problem solutions typically use optimization formulations that balance data fidelity and prior information. Selecting a suitable prior is critical to designing algorithms that converge efficiently and produce high-quality reconstructions. Traditional hand-crafted regularization priors can have limitations in handling complex and high-dimensional data encountered in signal, image, and video processing. The plug-and-play framework is a promising approach to solving inverse problems that replace traditional regularization priors with sophisticated denoisers or methods without being tied to an explicit optimization objective. Instead, the regularization prior is integrated as an operator in an iterative algorithm that alternates between data fidelity and prior steps. Advanced algorithms, like a denoiser or a deep neural network, can perform the prior step, offering several advantages, including bypassing the need for explicit inversion of the forward model and incorporating diverse and flexible methods.
Recent research advances have focused on developing adaptive plug-and-play algorithms that automatically adjust the regularization strength and parameters during the iterative process. Other works have explored the theoretical properties of the plug-and-play framework, such as convergence rates and stability guarantees.