Tada-DIP: Input-adaptive Deep Image Prior for One-shot 3D Image Reconstruction

TL;DR

Tada-DIP introduces an input-adaptive 3D Deep Image Prior method that enhances one-shot 3D image reconstruction quality, outperforming baselines and rivaling supervised approaches without requiring training data.

Contribution

The paper presents Tada-DIP, a novel fully 3D DIP technique combining input-adaptation and denoising regularization for improved one-shot 3D reconstruction.

Findings

Tada-DIP outperforms training-data-free baselines in 3D reconstruction quality.

Tada-DIP achieves performance comparable to supervised networks trained on large datasets.

The method effectively avoids overfitting common in traditional DIP approaches.

Abstract

Deep Image Prior (DIP) has recently emerged as a promising one-shot neural-network based image reconstruction method. However, DIP has seen limited application to 3D image reconstruction problems. In this work, we introduce Tada-DIP, a highly effective and fully 3D DIP method for solving 3D inverse problems. By combining input-adaptation and denoising regularization, Tada-DIP produces high-quality 3D reconstructions while avoiding the overfitting phenomenon that is common in DIP. Experiments on sparse-view X-ray computed tomography reconstruction validate the effectiveness of the proposed method, demonstrating that Tada-DIP produces much better reconstructions than training-data-free baselines and achieves reconstruction performance on par with a supervised network trained using a large dataset with fully-sampled volumes.