Search papers, labs, and topics across Lattice.
The paper introduces AnyBokeh, a physics-guided framework that enables any-to-any bokeh editing from a single image without the need for all-in-focus reconstruction. By estimating the source blur state through a signed circle-of-confusion map and a disparity map, AnyBokeh effectively transfers optical characteristics to achieve spatially adaptive deblurring and defocus rendering. Experiments demonstrate that this method outperforms existing techniques by providing faithful and controllable edits while eliminating the need for test-time bokeh-level calibration.
AnyBokeh revolutionizes bokeh editing by enabling precise control over depth-of-field effects directly from single images, sidestepping traditional reconstruction pitfalls.
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.
