一种基于时空域流形的视频序列图像拼接算法.pdfVIP

Proceedings of the 29th Chinese Control Conference July 29-31, 2010, Beijing, China 一种基于时空域流形的视频序列图像拼接算法* 1,2 1,2 1,2 1,2 邹丽晖 ，陈杰 ，张娟 ，窦丽华 1. 北京理工大学自动化学院, 北京100081 2. 复杂系统智能控制与决策教育部重点实验室, 北京100081 E-mail: chenji e@ 摘 要: 针对视频序列图像拼接自身具有帧间图像高重叠率高及序列数量多的特性，为更加合理利用视频序列图像帧 间重复信息，本文提出一种基于时空域流形的视频序列图像拼接算法，首先对视频序列各帧输入图像进行Gaussian 分层迭代计算Lucas-Kanade光流，得到较高精度的相邻帧间运动偏移量，对齐时空域序列图像，同时结合光流计算结 果消除前后帧间的亮度差异，然后对整体亮度均匀一致的视频序列图像采用时空域流形拼接技术合成高分辨率大视 场图像，实验表明本算法对视频序列图像拼接十分有效，拼接结果良好。 关键词:图像拼接, 时空域流形, 高斯分层, LK光流 An Image Mosaicing Approach for Video Sequences Based on * Space-Time Manifolds ZOU Li-Hui 1,2, CHEN Jie 1,2, ZHANG Juan 1,2, DOU Li-Hua 1,2 1. School of Automation, Beijing Institute of Technology, Beijing 100081, P.R.China 2. Key Laboratory of Complex System Intelligent Control and Decision (Beijing Institute of Technology), Ministry of Education, Beijing, 100081, P.R.China E-mail: chenji e@ Abstract: In order to utilize the repetitive information among the video sequence more reasonably, this paper proposes an image mosaicing approach based on space-time manifolds for the video sequence taking the properties of high-overlapped ratio between neighbor frames and multiple frames into account. Firstly, we iterate Lucas-Kanade optical flow under Gaussian pyramid strategy to compute the motion parameters between neighbor frames with higher accuracy and align the sequence, in the meantime, eliminate the brightness differenc