Boosting Monocular Metric Depth Estimation via Bokeh Rendering

Bokeh rendering and depth estimation share a fundamental optical connection, yet existing methods fail to fully exploit this reciprocity. Conventional bokeh pipelines rely heavily on noisy depth maps that inevitably introduce visual artifacts. Conversely, existing monocular depth models typically follow two flawed paradigms. Generative diffusion-based frameworks often lack consistent metric scale. Meanwhile, feed-forward metric depth models frequently fail in textureless or distant regions where defocus blur can provide geometric information. We propose BokehDepth, a two-stage framework that treats synthetic defocus as a supervision-free geometric signal. In the first stage, a physically grounded generative model produces calibrated bokeh stacks from a single sharp input without requiring prior depth input. Subsequently, a lightweight defocus-aware aggregation module integrates these stacks into the encoder of a depth estimation framework. This mechanism allows the model to extract consistent geometric features from the defocus dimension while keeping the decoder architecture unchanged. Experiments demonstrate that BokehDepth achieves superior visual bokeh fidelity compared to depth-dependent rendering baselines and consistently enhances the metric accuracy of state-of-the-art monocular depth models.

When we look at a photo, we can often sense which objects are close and which are far away. Teaching computers to recover this distance information from a single image is important for robotics, augmented reality, photography, and other vision applications, but it remains difficult, especially in flat, textureless, or distant regions. This paper explores a simple idea: the way a camera blurs out-of-focus areas, commonly known as bokeh, contains useful clues about depth. We introduce BokehDepth, a two-stage system that turns this visual clue into a practical tool. First, given a normal sharp image, our method generates several realistic versions of the image with different, controlled amounts of background blur, without needing a depth map beforehand. Then, a depth prediction model compares these images and learns how the scene changes as the blur strength varies. These changes help the model better infer which parts of the scene are near or far. Experiments show that our method produces more realistic bokeh than previous depth-based rendering pipelines and improves the accuracy of strong single-image depth estimation models across indoor and outdoor scenes. This suggests that camera blur, often treated only as an artistic effect, can also serve as a useful physical cue for understanding 3D structure from ordinary images.