Abstract:Non-line-of-sight (NLOS) imaging aims to reconstruct the three-dimensional hidden scenes from the data measured in the line-of-sight, which uses photon time-of-flight information encoded in light after multiple diffuse reflections. The under-sampled scanning data can facilitate fast imaging. However, the resulting reconstruction problem becomes a serious ill-posed inverse problem, the solution of which is highly possibility to be degraded due to noises and distortions. In this paper, we propose novel NLOS reconstruction models based on curvature regularization, i.e., the object-domain curvature regularization model and the dual (signal and object)-domain curvature regularization model. In what follows, we develop efficient optimization algorithms relying on the alternating direction method of multipliers (ADMM) with the backtracking stepsize rule, for which all solvers can be implemented on GPUs. We evaluate the proposed algorithms on both synthetic and real datasets, which achieve state-of-the-art performance, especially in the compressed sensing setting. Based on GPU computing, our algorithm is the most effective among iterative methods, balancing reconstruction quality and computational time. All our codes and data are available at <a class="link-external link-https" href="https://github.com/Duanlab123/CurvNLOS" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The paper primarily aims to address several key issues in Non-Line-Of-Sight (NLOS) imaging: 1. **The contradiction between the need for fast imaging and insufficient data**: To achieve high-speed imaging, researchers often use undersampling scanning strategies to reduce data acquisition time. However, this strategy leads to highly ill-posed reconstruction problems, where the reconstruction results may suffer from noise and distortion, resulting in degraded quality. 2. **The balance between the quality and speed of reconstruction algorithms**: Direct reconstruction methods, while fast, are sensitive to noise; iterative reconstruction methods can provide high-quality reconstructed images but require long computation times; deep learning-based methods rely on training datasets, and their performance on real measurement data may be affected. To address the above issues, the paper proposes the following contributions: - **A novel NLOS reconstruction model based on curvature regularization**: Utilizing curvature regularization to recover the smooth surfaces of hidden objects and fill in sparse measurement signals. This regularization method effectively ensures the smoothness of the reconstructed objects. - **Development of an efficient optimization algorithm**: A fast iterative optimization algorithm based on the Alternating Direction Method of Multipliers (ADMM), which can be implemented on a Graphics Processing Unit (GPU) to accelerate computation speed. - **Implementation of a GPU-accelerated NLOS reconstruction package**: Leveraging the parallel computing capabilities of GPUs, a reconstruction software package suitable for high-speed NLOS applications was developed. Through experiments on synthetic and real datasets, the paper demonstrates that the proposed curvature regularization model can effectively improve the quality of reconstructed images, especially in the case of undersampled data. Additionally, compared to existing techniques, this method significantly reduces computation time while maintaining high reconstruction quality.

Curvature regularization for Non-line-of-sight Imaging from Under-sampled Data

Statistical Image Reconstruction for Low-Dose Ct Using Nonlocal Means-Based Regularization

Non-line-of-sight reconstruction via structure sparsity regularization

Model-Guided Iterative Diffusion Sampling for NLOS Reconstruction

Domain Reduction Strategy for Non Line of Sight Imaging

Deep Non-line-of-sight Imaging from Under-scanning Measurements.

Dynamic-excitation-based Steady-State Non-Line-of-sight Imaging Via Multi-Branch Convolutional Neural Network

Towards Non-Line-of-Sight Photography

Multi-scale Iterative Model-guided Unfolding Network for NLOS Reconstruction.

Few-shot Non-line-of-sight Imaging with Signal-surface Collaborative Regularization

Non-line-of-sight Imaging with Arbitrary Illumination and Detection Pattern

Non-line-of-sight reconstruction with signal–object collaborative regularization

Nonconvex ${L_ {1/2}} $-Regularized Nonlocal Self-similarity Denoiser for Compressive Sensing based CT Reconstruction

Recent Advances on Non-Line-of-Sight Imaging: Conventional Physical Models, Deep Learning, and New Scenes

Physics to the Rescue: Deep Non-line-of-sight Reconstruction for High-speed Imaging

Passive Non-Line-of-Sight Imaging Using Optimal Transport

An Efficient Way for Active None-Line-of-Sight: End-to-End Learned Compressed NLOS Imaging.

Non-line-of-sight Sparse Aperture ISAR Imaging Via A Novel Detail-aware Regularization

Convex Latent-Optimized Adversarial Regularizers for Imaging Inverse Problems

Non-line-of-sight imaging based on Archimedean spiral scanning

Fast Non-Line-of-Sight Imaging With Hybrid Super-Resolution Network Over 18 m