TY - GEN
T1 - A novel video stabilization method with global motion vector prediction and border region recovery
AU - Zhang, Yang
AU - He, Xu
AU - Leng, Yuquan
AU - Liu, Chuanyin
N1 - Publisher Copyright:
© 2014 IEEE.
PY - 2014/4/20
Y1 - 2014/4/20
N2 - This paper presents a novel approach of vision stabilization, potentially applied to mobile robot or unmanned vehicle, which is inspired by the characteristics of human eye. Vision information processing is a key technology of intelligent mobile platform. However, in some condition, such as the vision module mount on the vehicle moving on the rugged ground, the system could not acquire stable vision, which influence the post-process of image information and make the observer uncomfortable. With video stabilization technology, the unwanted vibrations or shakes of video sequences is smoothed at real time speed. The algorithm proposed is mainly based on digital image stabilization technology, which divides the whole vision in 9 sub-area averagely, then processes searching and matching strategy by macro block with size of 16 X16 or 8×8 pixels. Both the start point for searching and motion vector are predicted by the motion trend. After the motion compensation procedure, the video is stabilized. During the procedure above, the edge area of the video is lost. In this paper, a recovery method inspired by human vision is discussed, which consider the duration of vision factors. With the evaluating a sequence of unstable images acquired by vision system on the mobile robot, we can draw the conclusion that the effectiveness and validation of the method aforementioned is implemented.
AB - This paper presents a novel approach of vision stabilization, potentially applied to mobile robot or unmanned vehicle, which is inspired by the characteristics of human eye. Vision information processing is a key technology of intelligent mobile platform. However, in some condition, such as the vision module mount on the vehicle moving on the rugged ground, the system could not acquire stable vision, which influence the post-process of image information and make the observer uncomfortable. With video stabilization technology, the unwanted vibrations or shakes of video sequences is smoothed at real time speed. The algorithm proposed is mainly based on digital image stabilization technology, which divides the whole vision in 9 sub-area averagely, then processes searching and matching strategy by macro block with size of 16 X16 or 8×8 pixels. Both the start point for searching and motion vector are predicted by the motion trend. After the motion compensation procedure, the video is stabilized. During the procedure above, the edge area of the video is lost. In this paper, a recovery method inspired by human vision is discussed, which consider the duration of vision factors. With the evaluating a sequence of unstable images acquired by vision system on the mobile robot, we can draw the conclusion that the effectiveness and validation of the method aforementioned is implemented.
KW - Digital video stabilization
KW - bioinspired
KW - duration of vision
KW - image matching
KW - motion estimation
UR - https://www.scopus.com/pages/publications/84949929129
U2 - 10.1109/ROBIO.2014.7090713
DO - 10.1109/ROBIO.2014.7090713
M3 - 会议稿件
AN - SCOPUS:84949929129
T3 - 2014 IEEE International Conference on Robotics and Biomimetics, IEEE ROBIO 2014
SP - 2483
EP - 2488
BT - 2014 IEEE International Conference on Robotics and Biomimetics, IEEE ROBIO 2014
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2014 IEEE International Conference on Robotics and Biomimetics, IEEE ROBIO 2014
Y2 - 5 December 2014 through 10 December 2014
ER -