Real-Time and High-Fidelity Non-Line-of-Sight Imaging

Non-line-of-sight (NLOS) imaging, which aims to reconstruct objects hidden from direct view, including see-through-the-medium and see-around-the-corner scenario categories, has become a promising field with broad applications. In this work, we introduce a unified NLOS reconstruction framework that addresses both categories of NLOS imaging problems. By incorporating scale modulation and joint regularization terms, the framework efficiently recovers albedo and depth across diverse measurement settings while enhancing reconstruction quality. To the best of our knowledge, this is the first method to deliver high-fidelity reconstruction with high computational efficiency across general NLOS imaging scenarios, providing a practical solution to real-world challenges. Moreover, we introduce a novel dataset that covers multiple measurement settings for both scenario categories, supporting future research in the field.