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

2025-07-13

AI Summary

Phase retrieval—reconstructing a signal from intensity-only measurements—is critical in imaging and crystallography, yet existing algorithms suffer from sensitivity to initialization and poor noise robustness. This paper proposes a deep iterative optimization framework grounded in image-to-image diffusion models. It first introduces a hybrid iterative initialization scheme integrating Hybrid Input-Output and Error Reduction, then employs a learned diffusion model for progressive refinement, augmented with acceleration mechanisms and feature aggregation to enhance stability. Within the Inversion by Direct Iteration framework, the method unifies prior knowledge and physical constraints in a single differentiable model. Experiments demonstrate that our approach significantly outperforms both classical and state-of-the-art deep methods in training efficiency and reconstruction accuracy, while exhibiting superior robustness to diverse noise types and sampling conditions, as well as enhanced generalization across unseen scenarios.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Phase retrieval involves recovering a signal from intensity-only measurements, crucial in many fields such as imaging, holography, optical computing, crystallography, and microscopy. Although there are several well-known phase retrieval algorithms, including classical iterative solvers, the reconstruction performance often remains sensitive to initialization and measurement noise. Recently, image-to-image diffusion models have gained traction in various image reconstruction tasks, yielding significant theoretical insights and practical breakthroughs. In this work, we introduce a novel phase retrieval approach based on an image-to-image diffusion framework called Inversion by Direct Iteration. Our method begins with an enhanced initialization stage that leverages a hybrid iterative technique, combining the Hybrid Input-Output and Error Reduction methods and incorporating a novel acceleration mechanism to obtain a robust crude estimate. Then, it iteratively refines this initial crude estimate using the learned image-to-image pipeline. Our method achieves substantial improvements in both training efficiency and reconstruction quality. Furthermore, our approach utilizes aggregation techniques to refine quality metrics and demonstrates superior results compared to both classical and contemporary techniques. This highlights its potential for effective and efficient phase retrieval across various applications.

Problem

• Recover signals from intensity-only noisy measurements
• Improve phase retrieval robustness and accuracy
• Enhance reconstruction quality using diffusion models

Innovation

• Hybrid iterative technique for robust initialization
• Learned image-to-image pipeline for refinement
• Aggregation techniques to enhance quality metrics