采用显著性分析与改进边缘方向直方图特征的红外与可见光图像配准  

Registration of infrared image and visible image based on saliency and EOH feature analysis

作　　者：徐军[1] 付天宇[2] 杨健[2] 丰苏[3] XU Jun FU Tian-yu YANG Jian FENG Su(State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University ,Wuhan 430079, China Beijing Engineering Research Center of Mixed Reality and Advanced Display, School of Optics and Electronics, Beijing Institute of Technology, Beijing 100081, China Information Center, General Administration of Quality Supervision Inspection and Quarantine of the PRC,Beijing 100088, China)

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]北京市混合现实与新型显示工程技术研究中心北京理工大学光电学院,北京100081 [3]国家质量监督检验检疫总局信息中心,北京100088

出　　处：《光学精密工程》2016年第11期2830-2840,共11页Optics and Precision Engineering

基　　金：中国博士后科学基金资助项目(No.2015M580962)

摘　　要：为了实现红外图像与可见光图像的信息融合,弥补单一模态图像的不足,提出了一种基于显著性分析与改进的边缘方向直方图EOH(Edge Orientation Histogram)特征的红外与可见光图像配准算法。该算法首先利用显著性分析技术找到可见光图像中的重要信息,得到显著性图;将其与可见光图像融合,实现可见光图像中重要信息的划分。然后,利用自适应FAST(Features from Accelerated Segment Test)算法,探测可见光与红外图像上的特征点;利用改进的EOH,描述特征点。最后,根据描述计算特征点的相似性,在可见光与红外图像上找出对应的特征点,实现红外与可见光图像的匹配。在3种不同情况下对红外与可见光图像数据进行了配准实验。结果表明:在红外图像与可见光图像采集条件相似情况下,特征点正确匹配率为96.55%,而在图像采集条件差异较大的情况下,特征点正确匹配率可达74.21%。该算法可实现红外与可见光图像的精确快速匹配,即使红外图像与可见光图像采集的角度与位置均存在较大差异的情况下,仍可以满足红外与可见光图像匹配对精度和稳定性的要求。To realize the information fusion of infrared and visible images and make up the deficiency of the single modality image,a new algorithm based on saliency and Edge Orientation Histogram（EOH）features was proposed.Firstly,the saliency analysis was used to find the important information of thevisible image and to obtain the saliency map.By fusing it with the visible image,the important information in the visible image was divided.Then,adaptive Features from Accelerated Segment Test（FAST）algorithm was employed in detecting feature points on the visible image and infrared image,and the improved EOH was used to describe the detected feature points.Finally,corresponding feature points were found by calculating the similarity of feature points in the visible and infrared images and the infrared and visible images were matched.An image matching experiments at three conditions were carried out,and the results indicate that when the collection conditions between the infrared and visible images are similar,the feature matching accuracy reaches 96.55%.When the difference of collection conditions between the infrared and visible images is large,the feature matching accuracy still can reach 74.21%.The algorithm realizes fast and accurate matching of infrared and visible images,and meets the requirements of image matching for accuracy and stability,especially under a collection condition that the infrared and visible images are bigger different.

关 键 词：图像配准 红外图像 可见光图像 显著性分析 自适应特征点提取 边缘方向直方图 特征描述 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]