LamiGauss: Pitching Radiative Gaussian for Sparse-View X-ray Laminography Reconstruction

2025-09-17

AI Summary

Poor reconstruction quality and severe artifacts plague sparse-view X-ray tomography of plate-like structures (e.g., microchips, composite batteries). To address this, we propose LamiGauss—a novel end-to-end sparse-projection optimization algorithm. LamiGauss integrates radiative Gaussian lattice rendering with precise tomographic geometric modeling, incorporating a differentiable detector-to-world coordinate transformation, explicit tilt-angle parameterization, and a pre-filtered initialization strategy. Compared to conventional iterative methods, LamiGauss achieves superior reconstruction accuracy and efficiency using only 3% of full-angle projection data, both on synthetic and real datasets. It effectively suppresses streaking and aliasing artifacts while preserving structural fidelity. The method establishes a new, efficient, and robust paradigm for 3D non-destructive evaluation of plate-like materials, enabling high-fidelity volumetric reconstruction from highly undersampled projections.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Matrix & Tensor Methods
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

X-ray Computed Laminography (CL) is essential for non-destructive inspection of plate-like structures in applications such as microchips and composite battery materials, where traditional computed tomography (CT) struggles due to geometric constraints. However, reconstructing high-quality volumes from laminographic projections remains challenging, particularly under highly sparse-view acquisition conditions. In this paper, we propose a reconstruction algorithm, namely LamiGauss, that combines Gaussian Splatting radiative rasterization with a dedicated detector-to-world transformation model incorporating the laminographic tilt angle. LamiGauss leverages an initialization strategy that explicitly filters out common laminographic artifacts from the preliminary reconstruction, preventing redundant Gaussians from being allocated to false structures and thereby concentrating model capacity on representing the genuine object. Our approach effectively optimizes directly from sparse projections, enabling accurate and efficient reconstruction with limited data. Extensive experiments on both synthetic and real datasets demonstrate the effectiveness and superiority of the proposed method over existing techniques. LamiGauss uses only 3$%$ of full views to achieve superior performance over the iterative method optimized on a full dataset.

Problem

• Reconstructing high-quality volumes from sparse-view X-ray laminography
• Overcoming geometric constraints in plate-like structure inspection
• Eliminating laminographic artifacts with optimized Gaussian splatting initialization

Innovation

• Gaussian Splatting radiative rasterization
• Detector-to-world transformation model
• Initialization strategy filtering laminographic artifacts