PIE-PS Photometric Stereo from Physical Irradiance Event Streams

Event cameras provide microsecond-level measurements and high dynamic range, but their sparse outputs depend on an unknown contrast threshold. PIE-PS learns to recover dense surface normals directly from raw events and known moving-light directions.

The framework builds physical irradiance observations from adjacent events, shares information across space and time, and learns to reduce the influence of unreliable measurements. Experiments on synthetic benchmarks and real captures consistently improve over prior event-based photometric stereo methods and a direct physics-only solver.