摘要
针对传统的尺度不变特征变换(SIFT)算法计算相对复杂、匹配耗时较长无法满足工程上实时计算要求的问题,提出一种基于信息熵的改进SIFT图像快速匹配算法。首先,根据局部熵自适应选择分割阈值把测试图像和参照图像分割成不均匀的两个区域,采用高熵区域做特征点检测,提高特征点的检测效率;然后利用每个子区域的梯度方向信息熵和梯度幅值信息熵把描述符向量的维度从128维降低到50维,降低欧氏距离的计算复杂度;最后,用最近邻距离比值法完成特征点匹配。通过实验对比发现改进的匹配算法在降低算法复杂度和提高正确匹配率的同时,显著地缩短了匹配时间。
For high algorithm complexity and long matching time of traditional scale invariant feature transform(SIFT) algorithm which cannot meet the real-time calculation requirements in engineering, an improved fast SIFT image matching algorithm based on information entropy is proposed. First, the matching image and the reference image are divided into two non-uniform regions by the adaptive segmentation threshold according to local entropy, and the high-entropy region is used for feature detection to improve the detection efficiency of feature points. Then the dimension of descriptor vector is reduced from 128 to 50 by using the gradient direction information entropy and gradient amplitude information entropy of each subregion, and the complexity of Euclidean distance calculation is reduced. Finally the nearest neighbor distance ratio method is used to complete the matching. Through experimental comparison, it is found that the improved matching algorithm reduces the complexity of the algorithm and improves the correct matching rate, while significantly shortening the matching time.
作者
刘自金
石玉英
LIU Zijin;SHI Yuying(School of Mathematics and Physics,North China Electric Power University,Beijing 102206,China)
出处
《激光杂志》
CAS
北大核心
2021年第12期129-135,共7页
Laser Journal
基金
北京市自然科学基金(No.Z200001)。
关键词
图像分割
图像匹配
SIFT特征
信息熵
image segmentation
image matching
SIFT feature
information entropy
