吴亮,郭俊峰,刘国英.高斯曲率耦合相关性制约规则的图像匹配算法[J].包装工程,2019,40(1):168-176.
WU Liang,GUO Jun-feng,LIU Guo-ying.An Image Matching Algorithm Based on Gauss Curvature and Correlation Constraint Rule[J].Packaging Engineering,2019,40(1):168-176. |
| 高斯曲率耦合相关性制约规则的图像匹配算法 |
| An Image Matching Algorithm Based on Gauss Curvature and Correlation Constraint Rule |
| 投稿时间：2018-08-31 修订日期：2019-01-10 |
| DOI：10.19554/j.cnki.1001-3563.2019.01.027 |
| 中文关键词: 图像匹配 高斯曲率模型 Hessian算子 灰度平均值 相关性制约规则 RANSAC算法 |
| 英文关键词: image matching Gauss curvature model Hessian operator gray mean value correlation constraint rule RANSAC algorithm |
| 基金项目:国家自然科学基金（41001251）；河南省重点科技攻关计划（102102310087）；河南省基础与前沿技术研究计划（152300410182） |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 针对当前较多图像匹配算法在匹配过程中因忽略了特征点之间的相关性而导致算法存在匹配正确度和鲁棒性不佳等不足，设计一种高斯曲率模型耦合相关性制约规则的图像匹配算法。方法 首先，利用高斯滤波后图像的一阶矩阵和Hessian矩阵来构造高斯曲率模型，对Hessian算子进行改进，以充分检测图像的特征点。然后，通过求取扇形区域内的Haar小波响应获取特征点的主方向，并根据特征点邻域中像素点的灰度平均值计算特征向量，从而形成特征描述子，完成对特征点的描述。利用特征点集的均值与协方差矩阵来构造相关性模型，对特征点的相关度完成度量，从而定义相关性制约规则，对特征点的相似度进行判断，完成特征点的匹配。最后，利用RANSAC算法提纯匹配特征点，完成图像的匹配。结果 仿真实验表明，与当前图像匹配算法相比较，文中算法不仅匹配正确度较高，且具有较强的鲁棒性，在旋转角度为50°时，其正确匹配精度仍可达到87%以上。结论 所提算法在多种几何攻击下仍具有较高的匹配精度，在图像处理、信息安全等领域具有良好的参考价值。 |
| 英文摘要: |
| The work aims to design an image matching algorithm based on Gauss curvature model and correlation constraint rule with respect to such deficiencies as low matching accuracy and robustness induced by ignoring the correlation between feature points in the matching process of many current image matching algorithms. First, the first-order matrix of the image after the Gauss filter and the Hessian matrix were used to construct the Gauss curvature model, and the Hessian operator was improved to fully detect the feature points of the image. Then, the main direction of the feature points was obtained by obtaining the Haar wavelet response in the sector area, and the eigenvectors were calculated according to the average gray-value of the pixel points in the neighborhood of the feature points, and thus the feature descriptors were formed to complete the description of feature points. The correlation model was constructed based on the mean and covariance matrix of the feature point set and the relevancy of feature points was measured. The similarity of the feature points was judged, and the matching of the feature points was completed. RANSAC algorithm was used to purify the matching feature points and complete the matching of images. The simulation results showed that, compared with the current image matching algorithm, the proposed algorithm not only had higher matching accuracy, but also had stronger robustness, whose correct matching accuracy could still reach over 87% when the rotation angle was 50°. The proposed algorithm still has higher matching accuracy under various geometric attacks, which has good reference value in image processing, information security and other fields. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
渝公网安备 50010702501716号