# Deep Inertia Lp Half-Quadratic Splitting Unrolling Network for Sparse View CT Reconstruction

arXiv: https://arxiv.org/abs/2408.06600 · HTML: https://mlsanity.com/paper/2408.06600/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 37 references that are in the map.
- First posted on arXiv 13 Aug 2024.

## Builds on (most cited first, 10 of 10)

- [Deep Convolutional Neural Network for Inverse Problems in Imaging](https://mlsanity.com/paper/1611.03679/) (2016)
- [Learned Primal-dual Reconstruction](https://mlsanity.com/paper/1707.06474/) (2017)
- [iPiano: Inertial Proximal Algorithm for Non-Convex Optimization](https://mlsanity.com/paper/1404.4805/) (2014)
- [MetaInv-Net: Meta Inversion Network for Sparse View CT Image Reconstruction](https://mlsanity.com/paper/2006.00171/) (2020)
- [Stage-by-stage Wavelet Optimization Refinement Diffusion Model for Sparse-View CT Reconstruction](https://mlsanity.com/paper/2308.15942/) (2023)
- [Convolutional Sparse Coding for Compressed Sensing CT Reconstruction](https://mlsanity.com/paper/1903.08549/) (2019)
- [Adaptive Graph-based Total Variation for Tomographic Reconstructions](https://mlsanity.com/paper/1610.00893/) (2016)
- [An efficient Quasi-Newton method for nonlinear inverse problems via learned singular values](https://mlsanity.com/paper/2012.07676/) (2020)
- [Visual Attention Network for Low Dose CT](https://mlsanity.com/paper/1810.13059/) (2018)
- [Fast, nonlocal and neural: a lightweight high quality solution to image denoising](https://mlsanity.com/paper/2403.03488/) (2024)

## Abstract (arXiv)

Sparse view computed tomography (CT) reconstruction poses a challenging ill-posed inverse problem, necessitating effective regularization techniques. In this letter, we employ Lp-norm (0<p<1) regularization to induce sparsity and introduce inertial steps, leading to the development of the inertial Lp-norm half-quadratic splitting algorithm. We rigorously prove the convergence of this algorithm. Furthermore, we leverage deep learning to initialize the conjugate gradient method, resulting in a deep unrolling network with theoretical guarantees. Our extensive numerical experiments demonstrate that our proposed algorithm surpasses existing methods, particularly excelling in fewer scanned views and complex noise conditions.