Source author record

Figen S. Oktem

Figen S. Oktem appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

eess.IV Computer Vision eess.SP physics.optics

Catalog footprint

What is connected

3works

4topics

2close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

I2I-PR: Deep Iterative Refinement for Phase Retrieval using Image-to-Image Diffusion Models

Phase retrieval aims to recover a signal from intensity-only measurements, a fundamental problem in many fields such as imaging, holography, optical computing, crystallography, and microscopy. Although there are several well-known phase retrieval algorithms, including classical alternating projection-based solvers, the reconstruction performance often remains sensitive to initialization and measurement noise. Recently, diffusion models have gained traction in various image reconstruction tasks, yielding significant theoretical insights and practical advances. In this work, we introduce a deep iterative refinement framework that redefines the role of diffusion models in phase retrieval. Instead of generating images from random noise, our method starts with multiple physically consistent initial estimates and iteratively refines them through a learned image-to-image diffusion process. This enables data-driven phase retrieval that is both interpretable and robust, leveraging the strengths of classical solvers while mitigating their weaknesses. Furthermore, we propose an enhanced initialization strategy that integrates classical algorithms with a novel acceleration mechanism to obtain reliable initial estimates. During inference, we adopt a geometric self-ensemble strategy based on input flipping, together with output aggregation to further improve the final reconstruction quality. Comprehensive experiments demonstrate that our approach achieves substantial gains in both training efficiency and reconstruction quality, consistently outperforming classical and recent state-of-the-art methods. These results highlight the potential of diffusion-driven refinement as an effective and general framework for robust phase retrieval across diverse applications. The source code and trained models are available at https://github.com/METU-SPACE-Lab/I2I-PR-for-Phase-Retrieval

preprint2026arXiv

prNet: Data-Driven Phase Retrieval via Stochastic Refinement

Phase retrieval is an ill-posed inverse problem in which classical and deep learning-based methods struggle to jointly achieve measurement fidelity and perceptual realism. We propose a novel framework for phase retrieval that leverages Langevin dynamics to enable efficient posterior sampling, yielding reconstructions that explicitly balance distortion and perceptual quality. Unlike conventional approaches that prioritize pixel-wise accuracy, our methods navigate the perception-distortion tradeoff through a principled combination of stochastic sampling, learned denoising, and model-based updates. The framework comprises three variants of increasing complexity, integrating theoretically grounded Langevin inference, adaptive noise schedule learning, parallel reconstruction sampling, and warm-start initialization from classical solvers. Extensive experiments demonstrate that our methods achieve state-of-the-art performance across multiple benchmarks, both in terms of fidelity and perceptual quality. The source code and trained models are available at https://github.com/METU-SPACE-Lab/prNet-for-Phase-Retrieval

preprint2020arXiv

Efficient Algorithms for Convolutional Inverse Problems in Multidimensional Imaging

Multidimensional imaging, capturing image data in more than two dimensions, has been an emerging field with diverse applications. Due to the limitation of two-dimensional detectors in obtaining the high-dimensional image data, computational imaging approaches have been developed to pass on some of the burden to a reconstruction algorithm. In various image reconstruction problems in multidimensional imaging, the measurements are in the form of superimposed convolutions. In this paper, we introduce a general framework for the solution of these problems, called here convolutional inverse problems, and develop fast image reconstruction algorithms with analysis and synthesis priors. These include sparsifying transforms, as well as convolutional or patch-based dictionaries that can adapt to correlations in different dimensions. The resulting optimization problems are solved via alternating direction method of multipliers with closed-form, efficient, and parallelizable update steps. To illustrate their utility and versatility, the developed algorithms are applied to three-dimensional image reconstruction problems in computational spectral imaging for cases with or without correlation along the third dimension. As the advent of multidimensional imaging modalities expands to perform sophisticated tasks, these algorithms are essential for fast iterative reconstruction in various large-scale problems.