摘要
在不同的图像之间寻找若干稳健的匹配点是许多计算机视觉算法有效应用的关键.SIFT(Scale Invariant Feature Transforms)算法已经证明了能够实现稳健匹配点提取的任务,然而SIFT算法提取的匹配点往往处于目标内部的某些均匀的区域,并不包括目标的边角点.考虑到角点往往包含了目标关键的结构信息,因此利用SIFT匹配点的稳健性结合角点检测,既能够提高角点匹配的稳健性又能够减小角点匹配搜索范围.匹配实验证明了该方法能够有效抵御噪声的干扰以及尺度和视角变化的影响,具备良好的匹配稳健性.
To find the robust matching points between different views on the same object is the key step in a com- puter vision system in practice. The popular SIFT （ Scale Invariant Feature Transforms） has showed its power for ro- bust feature points matching. However, the edge and corner points are not included in the SIFT matching points. Considering the fact that the corner points represent the structure information of an object, we hence propose a ro- bust approach for corner matching via SIFT. This new approach improves the robustness while decreases the compu- ting burden in matching. A series of experiments on corner points matching illustrate that the proposed approach is not sensitive to changes in view,rotation or scaZe,and robust tn nni~ inV,~,#＂
出处
《南京信息工程大学学报（自然科学版）》
CAS
2012年第6期550-554,共5页
Journal of Nanjing University of Information Science & Technology（Natural Science Edition）
基金
广东省科技计划资助项目(2010-B010700025)
