Res-U2Net: Untrained Deep Learning for Phase Retrieval and Image Reconstruction

Abstract

Conventional deep learning-based image reconstruction methods require a large amount of training data which can be hard to obtain in practice. Untrained deep learning methods overcome this limitation by training a network to invert a physical model of the image formation process. Here we present a novel untrained Res-U2Net model for phase retrieval. We use the extracted phase information to determine changes in an object's surface and generate a mesh representation of its 3D structure. We compare the performance of Res-U2Net phase retrieval against UNet and U2Net using images from the GDXRAY dataset.

arXiv comment16 pages, 8 figures, 4 Tables

References in corpus (15)

• U${}^2$-Net: Going Deeper with Nested U-Structure for Salient Object Detection
• U-Net and its variants for medical image segmentation: theory and applications
• Phase Retrieval via Wirtinger Flow: Theory and Algorithms
• Fringe pattern analysis using deep learning
• Image-based 3D Object Reconstruction: State-of-the-Art and Trends in the Deep Learning Era
• Experimental robustness of Fourier Ptychography phase retrieval algorithms
• On the use of deep learning for phase recovery
• Computational ghost imaging using deep learning
• Phase Retrieval: From Computational Imaging to Machine Learning
• PhaseGAN: A deep-learning phase-retrieval approach for unpaired datasets
• Phase Retrieval from 1D Fourier Measurements: Convexity, Uniqueness, and Algorithms
• Untrained networks for compressive lensless photography
• Deep Iterative Reconstruction for Phase Retrieval
• A Two-step-training Deep Learning Framework for Real-time Computational Imaging without Physics Priors
• The scaling and skewness of optimally transported meshes on the sphere