Dynamic Black-hole Emission Tomography with Physics-informed Neural Fields
Researchers have developed PI-DEF, a novel physics-informed neural rendering approach for dynamic 3D black hole imaging. This method improves upon previous techniques like BH-NeRF by not assuming Keplerian dynamics, which are inaccurate near black holes. PI-DEF jointly reconstructs the emissivity field and velocity field, enforcing the velocity as a soft constraint on emissivity dynamics, leading to more accurate reconstructions and the potential to estimate black hole physics parameters such as spin. AI