BokehDiff: Neural Lens Blur with One-Step Diffusion

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhu, Chengxuan, Fan, Qingnan, Zhang, Qi, Chen, Jinwei, Zhang, Huaqi, Xu, Chao, Shi, Boxin
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911219092291584
author Zhu, Chengxuan
Fan, Qingnan
Zhang, Qi
Chen, Jinwei
Zhang, Huaqi
Xu, Chao
Shi, Boxin
author_facet Zhu, Chengxuan
Fan, Qingnan
Zhang, Qi
Chen, Jinwei
Zhang, Huaqi
Xu, Chao
Shi, Boxin
contents We introduce BokehDiff, a novel lens blur rendering method that achieves physically accurate and visually appealing outcomes, with the help of generative diffusion prior. Previous methods are bounded by the accuracy of depth estimation, generating artifacts in depth discontinuities. Our method employs a physics-inspired self-attention module that aligns with the image formation process, incorporating depth-dependent circle of confusion constraint and self-occlusion effects. We adapt the diffusion model to the one-step inference scheme without introducing additional noise, and achieve results of high quality and fidelity. To address the lack of scalable paired data, we propose to synthesize photorealistic foregrounds with transparency with diffusion models, balancing authenticity and scene diversity.
format Preprint
id arxiv_https___arxiv_org_abs_2507_18060
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BokehDiff: Neural Lens Blur with One-Step Diffusion
Zhu, Chengxuan
Fan, Qingnan
Zhang, Qi
Chen, Jinwei
Zhang, Huaqi
Xu, Chao
Shi, Boxin
Computer Vision and Pattern Recognition
We introduce BokehDiff, a novel lens blur rendering method that achieves physically accurate and visually appealing outcomes, with the help of generative diffusion prior. Previous methods are bounded by the accuracy of depth estimation, generating artifacts in depth discontinuities. Our method employs a physics-inspired self-attention module that aligns with the image formation process, incorporating depth-dependent circle of confusion constraint and self-occlusion effects. We adapt the diffusion model to the one-step inference scheme without introducing additional noise, and achieve results of high quality and fidelity. To address the lack of scalable paired data, we propose to synthesize photorealistic foregrounds with transparency with diffusion models, balancing authenticity and scene diversity.
title BokehDiff: Neural Lens Blur with One-Step Diffusion
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2507.18060