모션 블러를 활용한 이미지·비디오 복원: 대규모 데이터 기반 딥러닝 접근

📝 Abstract

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. image details and degrades visual quality, it also encodes information about scene and camera motion during an exposure. Previous techniques leverage this information to estimate a sharp image from an input blurry one, or to predict a sequence of video frames showing what might have occurred at the moment of image capture. However, they rely on handcrafted priors or network architectures to resolve ambiguities in this inverse problem, and do not incorporate image and video priors on large-scale datasets. As such, existing methods struggle to reproduce complex scene dynamics and do not attempt to recover what occurred before or after an image was taken.

💡 Analysis

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. image details and degrades visual quality, it also encodes information about scene and camera motion during an exposure. Previous techniques leverage this information to estimate a sharp image from an input blurry one, or to predict a sequence of video frames showing what might have occurred at the moment of image capture. However, they rely on handcrafted priors or network architectures to resolve ambiguities in this inverse problem, and do not incorporate image and video priors on large-scale datasets. As such, existing methods struggle to reproduce complex scene dynamics and do not attempt to recover what occurred before or after an image was taken.

📄 Content

Generating the Past, Present and Future from a Motion-Blurred Image SAIKIRAN TEDLA, York University, Canada KELLY ZHU, University of Toronto, Canada and Vector Institute, Canada TREVOR CANHAM, York University, Canada FELIX TAUBNER, University of Toronto, Canada and Vector Institute, Canada MICHAEL S. BROWN, York University, Canada KIRIAKOS N. KUTULAKOS, University of Toronto, Canada and Vector Institute, Canada DAVID B. LINDELL, University of Toronto, Canada and Vector Institute, Canada (a) blurry in-the-wild photo (d) bringing motion-blurred historical photos to life (e) 4D scene reconstruction (c) tracking through past, present & future (b) sharp video frame from future present (motion-blurred) camera poses present future present past future 2D motion field past present motion-blurred present (motion-blurred) past present past present past present future future future future Fig. 1. (a) Given a motion-blurred input image, our approach uses a large-scale video diffusion model to generate frames that reveal scene motion during the exposure and predict what may have occurred just before and after the image was captured. We illustrate scene motion predicted by our method with (b) an output video frame and (c) tracking from an off-the-shelf method [Karaev et al. 2024]. The resulting videos capture complex scene dynamics, enabling downstream applications including (d) bringing historical images to life: we show insets of three sharp generated video frames (red, green, and blue bars indicate each frame’s exposure window) and visualize subtle motions revealed by the video as a 2D motion field computed by RAFT [Teed and Deng 2020]. (e) We can also recover dynamic 3D structure and camera poses by applying a recent structure from motion technique to our output video [Li et al. 2025]. Video results are included in the supplemental webpage. Photos: (top) © Thales Antônio, iStock; (bottom) U.S. National Archives and Records, public domain. We seek to answer the question: what can a motion-blurred image reveal about a scene’s past, present, and future? Although motion blur obscures Authors’ Contact Information: SaiKiran Tedla, tedlasai@yorku.ca, York University, Canada; Kelly Zhu, zhu@cs.toronto.edu, University of Toronto, Canada and Vector Institute, Canada; Trevor Canham, tcanham@yorku.ca, York University, Canada; Felix Taubner, ftaubner@cs.toronto.edu, University of Toronto, Canada and Vector Institute, Canada; Michael S. Brown, mbrown@eecs.yorku.ca, York University, Canada; Kiriakos N. Kutulakos, kyros@cs.toronto.edu, University of Toronto, Canada and Vector Institute, Canada; David B. Lindell, lindell@cs.toronto.edu, University of Toronto, Canada and Vector Institute, Canada. This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 Interna- tional License. image details and degrades visual quality, it also encodes information about scene and camera motion during an exposure. Previous techniques leverage this information to estimate a sharp image from an input blurry one, or to predict a sequence of video frames showing what might have occurred at the moment of image capture. However, they rely on handcrafted priors or network architectures to resolve ambiguities in this inverse problem, and do not incorporate image and video priors on large-scale datasets. As such, existing methods struggle to reproduce complex scene dynamics and do not attempt to recover what occurred before or after an image was taken. © 2025 Copyright held by the owner/author(s). ACM 1557-7368/2025/12-ART202 https://doi.org/10.1145/3763306 ACM Trans. Graph., Vol. 44, No. 6, Article 202. Publication date: December 2025. arXiv:2512.19817v1 [cs.CV] 22 Dec 2025 202:2 • SaiKiran Tedla, Kelly Zhu, Trevor Canham, Felix Taubner, Michael S. Brown, Kiriakos N. Kutulakos, and David B. Lindell Here, we introduce a new technique that repurposes a pre-trained video diffusion model trained on internet-scale datasets to recover videos revealing complex scene dynamics during the moment of capture and what might have occurred immediately into the past or future. Our approach is robust and versatile; it outperforms previous methods for this task, generalizes to challenging in-the-wild images, and supports downstream tasks such as recovering camera trajectories, object motion, and dynamic 3D scene structure. Code and data are available at blur2vid.github.io CCS Concepts: • Computing methodologies →Computer vision; Com- putational photography. Additional Key Words and Phrases: exposure control, deblurring, video diffusion model ACM Reference Format: SaiKiran Tedla, Kelly Zhu, Trevor Canham, Felix Taubner, Michael S. Brown, Kiriakos N. Kutulakos, and David B. Lindell. 2025. Generating the Past, Present and Future from a Motion-Blurred Image. ACM Trans. Graph. 44, 6, Article 202 (December 2025), 15 pages. https://doi.org/10.1145/3763306 1 Introduction “Only photography has been able to divide human life into a series of moments, each of them has the

This content is AI-processed based on ArXiv data.