Light Transport-aware Diffusion Posterior Sampling for Single-View Reconstruction of 3D Volumes

Figure 1 . Given a single view (y) of a volume (V ), we reconstruct a volume ( V ) from its latent representation (θ) that matches y under the same lighting conditions, resulting in a synthesized view (ŷ). A differentiable volume renderer (R) is used to optimize physical scene parameters (ϕ) while simultaneously performing posterior sampling p(θ|y; ϕ), conditioned on the observation, in the latent space of a trained diffusion model p(θ). Ambiguities due to the absence of information about unseen parts of the volume are reduced by gradually steering the reverse diffusion process toward the most plausible reconstruction under the given view (right section).