摘要
受分辨率和视角范围的限制,普通的图像获取设备通常无法拍摄到较大场景的图像,而图像拼接技术可对获取的若干幅小视角场景的图像进行配准和融合得到一幅宽视角的全景图像.为此,设计了一种序列图像拼接系统.针对近似仅存在二维平移的待拼接图像,采取平移鲁棒的实时序列图像拼接方案,以傅里叶变换的相位相关法为基础进行拼接.针对待拼接图像间同时存在旋转和平移的情况,采取了基于Harris特征点检测的旋转和平移鲁棒的序列图像拼接方案.实验结果表明,所设计的平移鲁棒序列图像拼接方案能够实时拼接近似存在二维平移关系的序列图像,但无法处理旋转;而旋转和平移鲁棒的序列图像拼接方案可在线下非实时地处理同时存在平移和旋转仿射变换的序列图像拼接.
Due to limitations in image resolution and viewing angles of the lens, ordinary image acquisition devices cannot capture images of super-large scenes. Image mosaicking techniques are used to produce panoramic images in such circumstances by registering and stitching a number of images with a limited system for this purpose. The first module perspective angle. We design an image mosaic is based on a phase correlation algorithm to realize real-time mosaicking of images with only translational relationship. The second module deals with images having both of rotational and translational relationships, which is based on Harris corner match. Experimental results show that the system can effectively achieve seamless image mosaicking.
出处
《应用科学学报》
CAS
CSCD
北大核心
2015年第4期376-388,共13页
Journal of Applied Sciences
基金
国家自然科学基金(No.61303203)
上海市自然科学基金(No.13ZR1428400)
上海市教育委员会科研创新项目基金(No.14YZ087)
上海智能家居大规模物联共性技术工程中心项目基金(No.GCZX14014)
沪江基金研究基地专项基金(No.C14001
No.C14002)资助
关键词
图像拼接
仿射变换
相位相关
Harris特征点
序列图像
image mosaicking, affine transform, phase correlation, Harris feature point, image sequence
