Nonconfocal non-line-of-sight imaging with specular-flight-path regularization for complex multi-orientation objects

Nonconfocal non-line-of-sight (NLOS) imaging, which typically relies on a single illumination point, often suffers from limited adaptability and low resolution in complex scenes. In this paper, we employ local matrix analysis to demonstrate that specular-flight-path regions exhibit geometric advantages of low column similarity and reduced information redundancy. Building on this, we construct a specular-flight-path regularization framework and design an associated optimization algorithm to enable complete and accurate reconstruction of hidden scenes. Extensive evaluations confirm that our method consistently achieves structural similarity index measure (SSIM) > 0.85 on simulated data and outperforms existing approaches on real data, especially in tilted multiobject scenarios. This work offers a new perspective to enhance the practicality and robustness of nonconfocal NLOS imaging.