摘要
针对传统互信息图像配准拼接算法计算量大、效率低等问题,本文结合模板匹配,提出基于模板与互信息的全景图拼接技术。首先将误差法和二次匹配误差法相结合,对待拼接图像进行初次模板匹配,划定大致重叠区域;接着从互信息量的角度比较相邻重叠的两幅图像的相似性,通过建立两幅图像之间的互信息量,计算最大互信息,获得匹配区域;然后再次利用模板匹配,设定最佳匹配区域,最终实现图像配准拼接。在VS2010+Opencv环境中编程实现重叠图像的拼接,并验证了算法的正确性。实验表明,本文算法具有计算量相对小,自动化程度高,配准拼接精度高等优点。
Traditional mutual information(MI) image registration stitching algorithm has the problems of large amount of calculation and low efficiency. So, this paper presents the mutual information(MI) and template matching panorama stitching techniques , based on template matching. Firstly, combining the error method and the second matching error method, the images ,which are gong to stitch, have on the initial template matching, and roughly de-fine the overlap areas; then from the perspective of mutual information(MI) ,compare the overlapped similarity of two neighboring images, through establishing the mutual information(MI) between two images, calculate the maxi-mum mutual information(MI), access to the matching region;then re-use template matching, set the best matching re-gion, and ultimately come true the registration and stitching of overlapped images. Matching and stitching the over-lapping images is in VS2010 and Opencv programming environment. At last, the experiment verify the correctness of the algorithm. Experimental results show that the proposed algorithm has a relatively small amount of calculation, a high degree of automation, and high registration stitching accuracy etc.
出处
《激光杂志》
CAS
CSCD
北大核心
2014年第9期62-65,共4页
Laser Journal
关键词
图像配准
模板匹配
互信息
全景图拼接
Iinage registration
Template matching
Mutual information(MI)
Panorama stitching