Differentiable Diffusion for Dense Depth Estimation from Multi-view Images

TL;DR

This paper introduces a differentiable diffusion approach for dense depth estimation from multi-view images, optimizing sparse points to produce accurate depth maps with improved robustness and efficiency.

Contribution

It proposes a novel differentiable diffusion method that optimizes sparse points for dense depth estimation, enabling high-quality reconstruction from multi-view images.

Findings

High-quality dense depth maps achieved from sparse points

Improved accuracy over traditional image processing and deep learning methods

Efficient optimization routine for large-scale scene reconstruction

Abstract

We present a method to estimate dense depth by optimizing a sparse set of points such that their diffusion into a depth map minimizes a multi-view reprojection error from RGB supervision. We optimize point positions, depths, and weights with respect to the loss by differential splatting that models points as Gaussians with analytic transmittance. Further, we develop an efficient optimization routine that can simultaneously optimize the 50k+ points required for complex scene reconstruction. We validate our routine using ground truth data and show high reconstruction quality. Then, we apply this to light field and wider baseline images via self supervision, and show improvements in both average and outlier error for depth maps diffused from inaccurate sparse points. Finally, we compare qualitative and quantitative results to image processing and deep learning methods. http://visual.cs.brown.edu/diffdiffdepth