Generative-AI- and Optical-Flow-Based Aspect Ratio Enhancement of Videos

被引:0
作者
Palczewski, Tomasz [1 ]
Rao, Anirudh [1 ]
Zhu, Yingnan [1 ]
机构
[1] Samsung Res Amer, AI Team Visual Display Lab, 665 Clyde Ave, Mountain View, CA 94039 USA
来源
2024 16TH INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND COMPUTING, ICMLC 2024 | 2024年
关键词
Gen-AI; optical-flow; aspect ratio enhancement; neural enhancement;
D O I
10.1145/3651671.3651681
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The global ultra-wide display market is growing rapidly due to widespread consumer adoption. However, existing content, especially on TV screens, is mainly designed for lower aspect ratios like 4:3. As TVs shift towards wider formats, there's a need to develop a method to transform legacy content and adapt to varying screen sizes. This shift holds the potential to revolutionize the content viewing experience for gamers, content creators, and streaming enthusiasts. While delivering high-quality visual content for dynamic aspect ratios remains a challenge, recent advancements in Deep Learning show promise in addressing these problems with generative approaches. Our contribution begins with a survey of prior video completion efforts, providing a foundational backdrop. We then introduce our novel solution, combining optical flow methodologies with generative latent diffusion models. These models, conditioned on an initial prompt and evolving video frame, refine content generation. We validate our approach on the DAVIS dataset, demonstrating its efficacy and robustness. In summary, our study pioneers advancements in content generation for ultra-wide displays.
引用
收藏
页码:355 / 362
页数:8
相关论文
共 21 条
[1]  
Blattmann A, 2023, Arxiv, DOI arXiv:2304.08818
[2]  
Gao C, 2020, Arxiv, DOI [arXiv:2009.01835, 10.48550/ARXIV.2009.01835]
[3]  
Goodfellow IJ, 2014, ADV NEUR IN, V27, P2672
[4]  
Heusel M, 2017, ADV NEUR IN, V30
[5]   Rethinking Image Inpainting via a Mutual Encoder-Decoder with Feature Equalizations [J].
Liu, Hongyu ;
Jiang, Bin ;
Song, Yibing ;
Huang, Wei ;
Yang, Chao .
COMPUTER VISION - ECCV 2020, PT II, 2020, 12347 :725-741
[6]   Globally and Locally Consistent Image Completion [J].
Iizuka, Satoshi ;
Simo-Serra, Edgar ;
Ishikawa, Hiroshi .
ACM TRANSACTIONS ON GRAPHICS, 2017, 36 (04)
[7]   A Survey on Data-Driven Video Completion [J].
Ilan, S. ;
Shamir, A. .
COMPUTER GRAPHICS FORUM, 2015, 34 (06) :60-85
[8]  
Lei CY, 2023, Arxiv, DOI arXiv:2303.08120
[9]  
Li JN, 2022, PR MACH LEARN RES
[10]   Context Encoders: Feature Learning by Inpainting [J].
Pathak, Deepak ;
Krahenbuhl, Philipp ;
Donahue, Jeff ;
Darrell, Trevor ;
Efros, Alexei A. .
2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2016, :2536-2544