【摘要】 针对多源传感器图像的特点,提出了一种结合高通能量变换与角点匹配的图像配准方法。该方法首先利用SUSAN算法进行角点检测,然后将角点匹配分为粗匹配和精匹配两个阶段实现:粗匹配时,对图像进行高通能量变换,有效减少多传感器图像间的差异,并利用高通能量图上角点邻域间的归一化互相关度量,来建立角点初始对应关系;精匹配时,通过RANSAC算法,进一步筛选出正确的匹配点对,并获取匹配参数。实验证明该方法能快速、准确地配准红外与可见光图像。更多还原

【Abstract】 This paper presents an effective multi-sensor image registration method combining high-pass energy transformation and feature point matching.First,the SUSAN algorithm is used to obtain the corners of the two images.And then,the matching procedure is realized with two stages:For coarse matching,the normalized local cross-correlation based on high-pass energy image is defined as similarity measurement to obtain the original correspondence;For precise matching,the robust RANSAC algorithm is applied to discard the incorrectly matched corner pairs and get the registrationt parameters.The validity and exactness have been shown by the experimental results.更多还原