融合显性和隐性度量的多模图像配准算法

为解决基于隐性度量的图像配准算法初始配准误差大容易引起失配的问题，本文提出了一种融合显性度量和隐性度量的多模图像分层配准算法。首先采用金字塔分解算法得到低分辨率待配准图像。然后在每一层金字塔图像中，先采用互信息作为显性度量，利用粒子群算法获得初始的配准参数；以此作为初始值，采用基于隐性度量的配准算法，利用融合粒子群和鲍威尔搜索法的优化算法获取更准确的配准参数。基于低分辨率图像计算得到配准参数后，先对高分辨率待配准图像进行变换，然后利用提出的上述算法进行参数优化，根据每层得到的配准参数计算最终的配准参数。最后，采用可见光与红外图像、多波段SAR图像进行了配准实验。实验结果表明，提出的算法适用于多模图像配准，能够减小配准误差，具有一定的适用性。 

Multimodal Image Registration Algorithm by Combining Explicit and Implicit Similarity

To solve the mismatch problem caused by big initial registration error in image registration algorithm based on implicit similarity, the algorithm of hierarchical combining registration algorithms based on explicit similarity and implicit similarity for multimodal image is proposed in this paper. Pyramid transform is first adopted to obtain low resolution images to be registrated. Mutual information is used as explicit similarity, and Particle Swarm Optimization (PSO) algorithm is adopted to compute the initial registration parameters. The algorithm based on implicit similarity is then used to obtain more accurate results by combination of PSO and Powell algorithm. The registration parameter above is used to transform the high resolution image to be registrated first, and registration parameter is then optimized using the method above, where the final registration parameter is computed using these two registration parameters. Several sets of visible images, visible and infrared images, and multi band SAR images are used to perform registration experiments. The experimental results demonstrate that, the proposed algorithm can be applied to multimodal image registration, reduce registration errors, and have certain applicability.