AShE – A Shadow Estimator for Augmented Reality Systems on Mobile Platforms

This work presents a geometric method for estimating the directional light vector in a scene, enabling realistic shadow generation in augmented reality through image segmentation and inverse rendering. The solution is designed for mobile devices, eliminating the need for specialized hardware and machine learning-based algorithms. The method employs standard cameras and fiducial markers to project realistic shadows on virtual objects, ensuring coherent visual integration with the real environment. The experiments demonstrated that the proposed approach achieved an average angular error of 11.42° for synthetic datasets, effectively estimating scene illumination and generating visually convincing shadow projections. Qualitative tests indicate that the system performs well under various lighting conditions, although it faces limitations in scenarios with translucent objects or diffuse lighting. The results suggest that this solution can serve as an efficient and accessible alternative for augmented reality applications on mobile devices, enhancing immersion and realism without requiring complex computational infrastructure.