🤖 AI Summary
Monocular depth estimation often suffers from over-smoothing, leading to a loss of geometric detail that limits scene understanding accuracy. This work proposes MDENeRF, a novel framework that, for the first time, incorporates uncertainty modeling from NeRF volume rendering into monocular depth refinement. By employing a Bayesian fusion strategy, the method iteratively integrates monocular priors with depth information derived from NeRF trained under multi-view perturbations, thereby preserving global structural consistency while recovering high-frequency geometric details. Evaluated on the SUN RGB-D indoor dataset, MDENeRF significantly improves depth estimation accuracy, producing depth maps with more accurate structures and richer fine-grained geometry.
📝 Abstract
Monocular depth estimation has applications in many fields, such as autonomous navigation and extended reality, making it an essential computer vision task. However, current methods often produce smooth depth maps that lack the fine geometric detail needed for accurate scene understanding. We propose MDENeRF, an iterative framework that refines monocular depth estimates using depth information from Neural Radiance Fields (NeRFs). MDENeRF consists of three components: (1) an initial monocular estimate for global structure, (2) a NeRF trained on perturbed viewpoints, with per-pixel uncertainty, and (3) Bayesian fusion of the noisy monocular and NeRF depths. We derive NeRF uncertainty from the volume rendering process to iteratively inject high-frequency fine details. Meanwhile, our monocular prior maintains global structure. We demonstrate improvements on key metrics and experiments using indoor scenes from the SUN RGB-D dataset.