Point Pairs Optimization for Piecewise Linear Transformation of Multimodal Remote Sensing Images by the Similarity of Log-Gabor Features

Abstract

The registration of multimodal remote sensing (MMRS) images often suffers from local deformations and obvious nonlinear radiometric differences. Piecewise linear (PL) transformation handles local deformations well but requires a stricter position accuracy of feature point pairs (FPPs). After outlier removal, the inevitably remaining outliers lead to lower local registration accuracy. To fix this issue, we propose a new similarity metric <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MSOLG_SSIM</i> , which combines the multi-scale and multi-orientation Log-Gabor features with structural similarity. The metric is then used to optimize the FPPs by maximizing the similarity between the fixed and the registered images to improve the registration accuracy of the PL transformation. Experimental results based on eight pairs of MMRS images demonstrated the robustness of the proposed similarity metric and the effectiveness of the FPPs optimization method. The code of the proposed method can be downloaded from https://github.com/HoucaiGuo/FPPs-Optimization-Multimodal.