Objective: Image registration is a important study direction of image process field and the basis of image fusion, image reconstruction and image analysis. Among the main image registration methods, methods based on image feature and methods based on image intensity have advantages and disadvantages respectively, combining the advantages of the two methods, we propose a new image registration method based on interesting points of rotation invariant feature. Methods: Firstly, with using Harris comer detector technology, the interest points on the model image and subject image could be extracted. Then the rotation invariant feature combined with intensity of each interest point were used as image features to construct a new cost function. Finally, the hierarchical optimization algorithm was used to optimize the cost function. At the beginning of image registration,we used the driving pixels to match each other to ensure the registration speed and robustness.then more and more points were used to registrate.so the registration accuracy was also good. Results: To show the advantage of our method, we use our method and method of B-spline based on mutual information to match the standard test images separately, the experimental results show that the registration accuracy of our method is better than B-spline based on mutual information.Con-clusions: Our method improves the registration accuracy with keeping the alignment robustness.%目的:图像配准是图像处理领域重要的研究方向,是图像融合、图像重建和图像分析等研究的基础.在图像配准的主要方法中,基于图像特征的配准方法和基于图像灰度的配准方法各有优缺点,通过结合这两种方法的优点,我们提出了一种基于感兴趣点的旋转不变性特征图像配准的新方法.方法:首先利用Harris角点检测技术,提取模板图像和目标图像的感兴趣点.然后把感兴趣点的旋转不变形特征和灰度值组成图像的特征描述向量,并提出新的代价函数.最后采用分级优化的策略优化代价函数,在配准初期,采用显著的特征点进行配准,以保证配准的速度与鲁棒性,随后通过逐步增加特征点的数量,则保证了配准的精度.结果:为显示本文方法的优越性,实验利用本文方法和基于互信息的B样条方法分别对标准测试图像进行配准,实验结果表明,本文方法较基于互信息的B样条方法在配准精度上有明显提高.结论:本文方法在保持配准鲁棒性的前提下,获得了较高的配准精度.
展开▼
