Fast image registration method based on Harris and SIFT algorithm
-
摘要: 本文提出了一种结合Harris与SIFT算子的快速图像配准方法。首先,对Harris算法进行两方面的改进:一是构建高斯尺度空间,提取具有尺度不变性的角点特征;二是采用Forsnter算子对提取的角点精定位,提高配准精度。然后,利用SIFT算子的特征描述方法描述提取到的特征点,通过随机kd树算法对两幅影像的特征点进行匹配。最后采用RANSAC算法对匹配点对进行提纯,并通过最小二乘法估计两幅影像间的空间变换单应矩阵,完成图像配准。实验结果表明:本文方法在基本保持配准精度的同时,在配准过程的时间消耗上比标准SIFT算法减少了64%。
-
关键词:
- 图像配准 /
- 多尺度Harris算子 /
- SIFT /
- RANSAC
Abstract: A new method for fast image registration based on improved Harris-Sift algorithm is proposed. Firstly, classic Harris algorithm is improved by building Gaussian scale space to extract scale invariant Harris corners and they are refined to sub-pixel corners using Forsnter algorithm. Then the SIFT descriptor is utilized to characterize those feature points and the matching procedure is carried out via randomized kd trees. At last, RANSAC is used to remove wrong matches and the optimal transform parameters are estimated using the least square method to accomplish the image registration process. The experimental results demonstrate that compared with the classic SIFT algorithm the proposed method decreases the cost time of the registration procedure mostly by 64% while almost keeping the same performance.-
Key words:
- image registration /
- multiple scale Harris operator /
- SIFT /
- RANSAC
-
图 5 图3(c)和图3(d)匹配结果及配准结果
Figure 5. Results of (a)feature matching and (b)registration for Fig.3(c) and Fig.3(d)
表 1 分别用本文方法以及SIFT、SURF进行特征点检测的时间统计
Table 1. Statistics of cost time for feature points detection using SIFT, SURF and the proposed method
Image Number of feature points Time of feature points detection/s SIFT SURF Harris-SIFT SIFT SURF Harris-Sift Fig.3(a) 3 857 770 1 065 9.32 4.47 4.91 Fig.3(c) 641 199 297 1.72 0.89 0.91 Fig.3(e) 2 938 1 856 1 835 1.39 0.44 0.29 
下载: 导出CSV
表 2 本文方法特征点匹配的正确率和时间统计
Table 2. Statistics of accuracy and cost time of feature points matching using the proposed method
Image Number of feature points Matches Correct matches Percent of correct matches/% Total time/s (a) 1 065 197 155 78.68 10.31 (b) 1 146 (c) 297 130 113 86.92 1.93 (d) 304 (e) 1 835 152 129 84.87 3.15 (f) 1 856
下载: 导出CSV
表 3 SIFT算法特征点匹配的正确率和时间统计
Table 3. Statistics of accuracy and cost time of feature points matching using SIFT
Image Number of feature points Matches Correct matches Percent of correct matches/% Total time/s (a) 3 857 701 495 70.61 30.45 (b) 4 007 (c) 641 296 244 82.43 5.34 (d) 654 (e) 2 938 236 197 83.47 11.80 (f) 3 643
下载: 导出CSV
表 4 SURF算法特征点匹配的正确率和时间统计
Table 4. Statistics of accuracy and cost time of feature points matching using SIFT
Image Number of feature points Matches Correct matches Percent of correct matches/% Total time/s (a) 770 50 37 74.00 8.27 (b) 600 (c) 199 65 60 92.31 1.53 (d) 278 (e) 1 856 127 113 88.98 3.64 (f) 1 847
下载: 导出CSV
表 5 使用不同方法对图3(c)和图3(d)进行配准的精度对比
Table 5. Comparison of registration accuracy for Fig.3(c) and Fig.3(d) using different methods
Method Percent of correct matches/% RMSE Time of image registration/s Harris-SIFT 86.92 0.47 1.93 SIFT 82.43 0.54 5.34
下载: 导出CSV
-
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] -
下载:
计量
- 文章访问数: 2671
- HTML全文浏览量: 839
- PDF下载量: 759
- 被引次数: 0
