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

发布时间：2018-06-10 17:49

本文选题：图像配准 + 红外图像　；参考：《光学精密工程》2016年11期

【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。
[Abstract]:In order to realize the information fusion between infrared image and visible image and to make up for the deficiency of single mode image, a registration algorithm of infrared and visible light image based on the feature of edge direction histogram (EOH) and edge direction histogram (EOHZ) is proposed. The algorithm firstly finds the important information in the visible image by using the saliency analysis technique, and obtains the salient image, and fuses it with the visible light image to realize the partition of the important information in the visible light image. Then, the feature points in the visible and infrared images are detected by the adaptive FASTX Features from Accelerated Segment Test algorithm, and the feature points are described by the improved EOH algorithm. Finally, according to the similarity of the feature points described, the corresponding feature points are found on the visible and infrared images, and the matching between infrared and visible images is realized. The registration experiments of infrared and visible image data were carried out in three different cases. The results show that the correct matching rate of feature points is 96.55 when the acquisition conditions of infrared images are similar to those of visible images, but the correct matching rate of feature points can reach 74.21 when the image acquisition conditions are quite different. The algorithm can match infrared and visible images accurately and quickly, even if the angle and position of infrared and visible images are different. It can still meet the accuracy and stability requirements of infrared and visible image matching.
【作者单位】：武汉大学测绘遥感信息工程国家重点实验室;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;国家质量监督检验检疫总局信息中心;
【基金】：中国博士后科学基金资助项目(No.2015M580962)
【分类号】：TP391.41

【相似文献】