一种基于显著性区域检测的特征匹配算法  

The improvement on speeded up robust features for images matching based on salient region detection

作　　者：康璐[1,2] 赵汝进[1] 任国强[1] 洪裕珍 KANG Lu ZHAO Ru-jin REN Guo-qiang HONG Yu-zhen(Institute of Optics and Electronics of Chinese Academy of Sciences, Chengdu 610209, China University of Chinese Academy of Sciences, Beijing 100149,China)

机构地区：[1]中国科学院光电技术研究所,四川成都610209 [2]中国科学院大学,北京100149

出　　处：《电子设计工程》2017年第2期136-139,共4页Electronic Design Engineering

基　　金：国家自然科学基金(61501429);中科院青年创新促进会(2016335)

摘　　要：针对传统特征匹配算法在实际的应用中存在搜索范围广、无关特征点多等问题,提出一种基于显著性区域检测特征匹配方法。首先利用显著性区域检测算法滤除图像中的背景,获取图像显著性区域;在此基础上利用SURF(Speed-up robust features,SURF)算法在显著区域内进行特征匹配;最后利用RANSAC(Random Sample Consensus,RANSAC)算法剔除误匹配,以匹配准确性。仿真验证试验中,图像特征匹配准确度平均提高7%左右,试验结果表明:基于显著性区域检测的特征匹配算法,能有效地降低图像背景对图像匹配的干扰,提高匹配的精度。In practice, many matching algorithmshave some problems that the searching range is very big and have many irrelevant feature points. This paper proposes a new algorithm which based on salient region detection to focus on those problems. At first, the new algorithm delete the background of image using the algorithm of salient region detection, and got the region of saliency. Then the paper computed the SURF features and matched it, usedRANSAC algorithms to remove the false pairs to improve the accuracy, finally. In the simulation experiments, the accuracy of matching had improve about 7%. From the experimental results, the new algorithm can reduce the interference of background for matching, and are more stable and more accurate for matching.

关 键 词：图像匹配 显著性区域检测 加速鲁棒特征 RANSAC 

分 类 号：TN911.73[电子电信—通信与信息系统]