摘要
鱼眼图片相对于普通图片来说具有更宽广的视域,使用2～4张鱼眼图片即可覆盖水平和垂直方向360度的场景.然而鱼眼图片中的场景呈现扭曲现象,因而不能直接使用一般的拼接方式合成全景图.笔者先使用一种鱼眼图片扭曲校正的方法并与相位相关度方法结合获取图像间偏移量的初值,然后迭代求解图像间亮度差的最小平方和函数,来求精图像间的偏移量以及图像扭曲校正参数.
Fisheye image has wider view lens compared to ordinary image. Two to four fisheye images are enough to cover 360-degree horizontal and vertical field of view . However, ordinary registration method causes bad panoramic image because of serious distortion of fisheye images. In this paper, a distortion correcting method is used and combined with phase correlation method to get the initial rotation values of the all image pairs with overlap. Then the least gray value difference square sum function of the all image pairs with overlap is iteratively solved to get precise rotation value and distortion correction parameter.
出处
《云南民族大学学报（自然科学版）》
CAS
2004年第3期214-217,共4页
Journal of Yunnan Minzu University:Natural Sciences Edition
基金
省院省校科技合作项目(2001JAALA02A022)
云南省自然科学基金资助项目(2000YP20)
云南省科技攻关重点资助项目(2002GG15)
关键词
扭曲校正
相位相关度
最小平方和函数
全景合成
distortion correction, phase correlation, least square sum function, panorama mosaic
