AnyBokeh: Physics-Guided Any-to-Any Bokeh Editing with Optical Fingerprint Transfer

2026-06-30

AI Summary

Existing depth-of-field editing methods rely on all-in-focus image reconstruction, often discarding original blur cues and introducing artifacts. This work proposes AnyBokeh, the first framework enabling arbitrary-to-arbitrary depth-of-field editing without requiring all-in-focus reconstruction or aperture calibration at test time. By estimating signed circle-of-confusion maps and disparity, AnyBokeh models their linear relationship to extract an optical fingerprint, which is then transferred to target focus and aperture settings—treating source image blur as a physically meaningful cue. Integrating physics-guided modeling, optical fingerprint transfer, and conditional generative blur synthesis, and trained on a high-fidelity synthetic dataset with accurate depth, focus distance, and EXIF metadata, AnyBokeh achieves high-fidelity, controllable depth-of-field re-rendering, bokeh synthesis, and defocus deblurring in real-world scenes, outperforming existing approaches that necessitate reconstruction or calibration.

Abstract

Depth-of-field control is a fundamental tool in photography, yet post-capture bokeh editing from a single image remains challenging. A practical editor should handle images captured under arbitrary focus and aperture settings. Existing methods typically assume an all-in-focus input, or first recover an all-in-focus image before rendering new bokeh. Such pipelines can discard useful blur cues from the source image and propagate reconstruction artifacts into the final edit. We introduce AnyBokeh, a physics-guided framework for any-to-any bokeh editing. Instead of treating source blur merely as a degradation to be removed, AnyBokeh estimates the source blur state with a signed circle-of-confusion map and a disparity map. By modeling the linear relation between signed circle of confusion and disparity difference, AnyBokeh estimates a source-specific optical fingerprint and transfers the source optical characteristics to the desired focus and aperture setting. A generative editor conditioned on both source and target circle-of-confusion maps then performs relative blur synthesis, enabling spatially adaptive deblurring, preservation, and defocus rendering. To support physically supervised learning, we further construct a high-fidelity synthetic dataset with accurate depth, focus distance, and full EXIF metadata. Experiments on real-world benchmarks show that AnyBokeh achieves faithful and controllable editing across any-to-any bokeh editing, all-in-focus-to-bokeh rendering, and defocus deblurring, while avoiding all-in-focus reconstruction and test-time bokeh-level calibration commonly required by existing approaches. The code and dataset will be available at https://github.com/itsmag11/AnyBokeh.

Problem

• bokeh editing
• depth-of-field control
• single-image defocus
• any-to-any editing
• optical fingerprint

Innovation

• any-to-any bokeh editing
• optical fingerprint transfer
• signed circle-of-confusion
• physics-guided generative editing
• relative blur synthesis