Robust Registration of Optical and SAR Images Using Multi-Orientation Relative Total Variation Structural Representation

Abstract

Accurate registration of optical and synthetic aperture radar (SAR) images remains a challenging task because of the potential large modality differences across individual images. To improve the registration performance, this article proposes a robust registration method for optical and SAR images based on a novel multi-orientation relative total variation (MORTV) structural representation. The MORTV model is designed by integrating multiple orientation strategy into the original RTV to extract the structural maps, which can capture more structural features while removing image noises and textures. Then, a novel feature descriptor called layerwise multiscale histogram of oriented gradient (LMHOG) is constructed on the multiscale structural maps that are generated using the MORTV model with different parameters. The LMHOG can fully characterize structural features at different scales in a multilayer manner, further enhancing the robustness and distinctiveness of the descriptor without increasing its dimension. Comprehensive experiments on two large-scale optical and SAR image datasets validate that the proposed method obtains superior registration performance over several state-of-the-art methods.

Keywords

• Image registration
• multimodal remote sensing images
• multi-orientation
• relative total variation (RTV)
• structural feature extraction