Robust Registration of Multimodal Remote Sensing Images With Spectrum Congruency

Among the existing registration methods, most feature descriptors are designed with image intensity, gradient information and phase congruency (PC). However, both intensity sensitive to illumination changes, complex differences, noise, etc. Despite fact that PC is invariant contrast, it does not perform well when images corrupted noise nonlinear radiation distortions. In this article, we propose a novel called spectrum (SC), which robust variations of intensity. SC focuses on exploiting correlation multiscale patches based their local energy measures distribution in data-driven transform domain. To demonstrate superiority SC, apply multimodal registration. We construct histogram-based descriptor termed as HOSC. Then HOSC integrated two similarity metrics for remote sensing Extensive experimental results real noisy pairs show proposed method presents superior accuracy excellent performance resisting distortion noise.