The increasing prevalence of infrared imagery in 3D reconstruction highlights the ongoing need to address challenges like low contrast and sparse texture, which significantly impact reconstruction accuracy. This paper introduces a framework for large-scale infrared 3D reconstruction, integrating Structure-from-Motion (SfM) optimization strategies with 3D Gaussian Splatting (3DGS). Initially, the Laplacian variance is used to assess and filter blurred images, improving the quality of the input dataset. Subsequently, an SfM-based optimized feature extraction and matching strategy is implemented, specifically designed to overcome the challenges inherent in infrared imagery, thereby enhancing the stability and robustness of feature points. Finally, the 3DGS method is applied, representing 3D points as Gaussian distributions to further enhance point cloud density and accuracy. Experimental results demonstrate that the proposed optimizations not only improve rendering quality but also eliminate redundant and erroneous matches, leading to a 35% increase in reconstruction efficiency.
Liu et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: