| Title |
Deep Learning-based Real-time Video Stabilization for Infrared Videos |
| Authors |
윤성준(Sungjoon Yoon) ; 심규진(Kyujin Shim) ; 고강욱(Kangwook Ko) ; 하남구(Namkoo Ha) ; 이민석(Minseok Lee) ; 장현성(Hyunsung Jang) ; 권구용(Kuyong Kwon) ; 박가영(Kayoung Park) ; 김창익(Changick Kim) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.2.77 |
| Keywords |
Video stabilization; Deep learning; Infrared video; Real-time processing; Acceleration |
| Abstract |
A wide range of research topics has been studied to overcome the physical limitations of cameras to enhance video quality. Among those topics, video stabilization aims to generate stable videos by removing annoying camera shakes from videos, and is an essential research field in obtaining high-quality videos. In this paper, we propose two deep learning-based video stabilization models with real-time performance that reduce small physical camera shakes from unstable infrared videos (e.g., videos taken from vehicles or small flying objects), which are widely used in low-light conditions. Our proposed models are trained in a self-supervised way, with no paired datasets consisting of unstable videos and their corresponding ground truth stable videos. The proposed stabilization models outperform the existing deep learning-based online stabilization model quantitatively and qualitatively, on infrared videos of 640x512 resolution with real-time (>30 fps) performance. |