BokehDepth: Enhancing Monocular Depth Estimation through Bokeh Generation

TL;DR

BokehDepth introduces a two-stage framework that improves monocular depth estimation by decoupling bokeh synthesis from depth prediction and leveraging defocus cues as auxiliary information.

Contribution

It presents a novel two-stage approach that separates bokeh generation from depth estimation and uses defocus as an auxiliary cue to enhance monocular depth accuracy.

Findings

Improves visual fidelity of depth-based bokeh rendering.

Boosts accuracy and robustness of existing monocular depth models.

Achieves better performance on challenging benchmarks.

Abstract

Bokeh and monocular depth estimation are tightly coupled through the same lens imaging geometry, yet current methods exploit this connection in incomplete ways. High-quality bokeh rendering pipelines typically depend on noisy depth maps, which amplify estimation errors into visible artifacts, while modern monocular metric depth models still struggle on weakly textured, distant and geometrically ambiguous regions where defocus cues are most informative. We introduce BokehDepth, a two-stage framework that decouples bokeh synthesis from depth prediction and treats defocus as an auxiliary supervision-free geometric cue. In Stage-1, a physically guided controllable bokeh generator, built on a powerful pretrained image editing backbone, produces depth-free bokeh stacks with calibrated bokeh strength from a single sharp input. In Stage-2, a lightweight defocus-aware aggregation module plugs into existing monocular depth encoders, fuses features along the defocus dimension, and exposes stable depth-sensitive variations while leaving downstream decoder unchanged. Across challenging benchmarks, BokehDepth improves visual fidelity over depth-map-based bokeh baselines and consistently boosts the metric accuracy and robustness of strong monocular depth foundation models.