为了解决多目偏振相机在获取图像中需要进行准确配准的问题, 本文提出了一种基于子图像的相位相关算法。算法首先对图像进行噪声减弱与细节增强, 并进行粗配准。之后在将图像分割为多个子图, 对每个对应区域的子图进行配准, 并剔除配准误差较大的子图, 以减少它们对于最终配准结果的不良影响。最后根据这些子图的配准参数来计算单应性矩阵, 完成最终的配准。实验结果表明, 对于3种不同场景的图像, 相比于传统方法, 本文方法的配准结果归一化互信息值均有所提高, 至多提高0.067%且图像细节更清晰。与传统算法相比, 本文算法对于低清晰度、大噪声的图像具有更好的配准效果。

To solve the problem that multi-lens polarization cameras need accurate registration in acquiring images, this paper proposes a phase correlation algorithm based on sub-image. The algorithm first performs noise reduction and detail enhancement on the image followed by a coarse registration. Then the image is divided into multiple subgraphs, the subgraph of each corresponding area is registered, and the subgraphs with large registration errors is eliminated to reduce their adverse effects on the final result. Finally the univariate matrix is calculated according to the registration parameters of the remaining subgraphs to complete the final registration. The experimental results show that for all three different types of images, the proposed algorithm improves both the normalized mutual information values, up to 0.067%, and the image details. Compared with the original algorithm, the proposed algorithm has a better registration effect for low definition and high noise images.