Digital image stabilization based on adaptive motion filtering with feedback correction
详细信息 查看全文
- 作者：Bo Zhai ; Jin Zheng ; Bo Li
- 关键词：Online digital image stabilization ; Motion filtering ; Kalman filter ; Feedback correction ; Objective evaluation
- 刊名：Multimedia Tools and Applications
- 出版年：2016
- 出版时间：October 2016
- 年：2016
- 卷：75
- 期：19
- 页码：12173-12200
- 全文大小：3,526 KB
- 刊物类别：Computer Science
- 刊物主题：Multimedia Information Systems
Computer Communication Networks
Data Structures, Cryptology and Information Theory
Special Purpose and Application-Based Systems - 出版者：Springer Netherlands
- ISSN：1573-7721
- 卷排序：75
文摘
For online digital image stabilization system, the camera usually moves with diverse and variable modes, which make the motion filtering process difficult to reserve the intentional fluctuations and remove unwanted jitters simultaneously. This paper presents an adaptive motion filtering algorithm with feedback correction. Firstly, based on the low frequency character of intentional motion in adjacent frames, the intentional velocity is regarded as the control variable, thus the modified one dimension Kalman filtering algorithm is proved to converge to a balance state of consistency and stabilization. Secondly, according to the mutual restraint of consistency and stabilization, this paper proposes two corresponding online feedback factors to reflect the immediate filtering performances. Hence, a motion filtering algorithm with improved Kalman filtering and parameter self-adjustment is realized, which can describe the real camera motion flexibly, as well as adapt to its changes. At last, an objective evaluation method for motion filtering is presented from the aspects of integral consistency, integral stabilization and integral robustness. Compared with other classical motion filtering algorithms, experimental results indicate that the proposed algorithm is more fast-computing and adaptive for different moving modes of the camera, which can reserve the intentional motions and remove the jitters steadily.