Algorithmic Guarantees for Inverse Imaging with Untrained Network Priors

Abstract

Deep neural networks as image priors have been recently introduced for problems such as denoising, super-resolution and inpainting with promising performance gains over hand-crafted image priors such as sparsity and low-rank. Unlike learned generative priors they do not require any training over large datasets. However, few theoretical guarantees exist in the scope of using untrained neural network priors for inverse imaging problems. We explore new applications and theory for untrained neural network priors. Specifically, we consider the problem of solving linear inverse problems, such as compressive sensing, as well as non-linear problems, such as compressive phase retrieval. We model images to lie in the range of an untrained deep generative network with a fixed seed. We further present a projected gradient descent scheme that can be used for both compressive sensing and phase retrieval and provide rigorous theoretical guarantees for its convergence. We also show both theoretically as well as empirically that with deep network priors, one can achieve better compression rates for the same image quality compared to hand crafted priors.

arXiv commentNeurIPS 2019 version with few modifications

References in corpus (10)

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Compressed Sensing using Generative Models
• Deep Decoder: Concise Image Representations from Untrained Non-convolutional Networks
• Compressed Sensing with Deep Image Prior and Learned Regularization
• Multi-Layer Convolutional Sparse Modeling: Pursuit and Dictionary Learning
• Towards moderate overparameterization: global convergence guarantees for training shallow neural networks
• Deep Compressed Sensing
• Regularizing linear inverse problems with convolutional neural networks
• Speech Denoising by Accumulating Per-Frequency Modeling Fluctuations
• Alternating Phase Projected Gradient Descent with Generative Priors for Solving Compressive Phase Retrieval

Cited by in corpus (11)

• 2022 Review of Data-Driven Plasma Science
• Deep Learning Methods for Solving Linear Inverse Problems: Research Directions and Paradigms
• Denoising and Regularization via Exploiting the Structural Bias of Convolutional Generators
• Fixed-Point and Objective Convergence of Plug-and-Play Algorithms
• Plug-and-Play Regularization using Linear Solvers
• Regularizing linear inverse problems with convolutional neural networks
• Self-Validation: Early Stopping for Single-Instance Deep Generative Priors
• Deep Learning Techniques for Inverse Problems in Imaging
• Unrolled Wirtinger Flow with Deep Decoding Priors for Phaseless Imaging
• Provably Convergent Algorithms for Solving Inverse Problems Using Generative Models
• Compressed Sensing via Measurement-Conditional Generative Models