Robust Descriptor Algorithm Considering the Changing Gray Value Trends Inside Ground Objects for Heterogeneous Optical Image Matching

Differences in sensor types, resolutions, and imaging conditions can lead to considerable spectral differences in heterogeneous optical remote sensing images and the similarity of scale-invariant feature transform (SIFT) or local self-similarities (LSS) feature descriptors of the same point can be poor. Consequently, we proposed a robust descriptor construction algorithm considering the changing gray values inside ground objects. The main contributions of this article include the following. First, based on the stability of the internal gray value changes of ground objects, we suggest that the change orientations and degrees of gray values of pixels can be used to express the stability of the same area of heterogeneous images, providing the basis for image matching; second, unlike many existing methods that use gradient information to calculate feature orientation and descriptors, the proposed algorithm uses change orientation and degree to calculate the feature orientation and descriptor, enabling it to obtain stable descriptors in image matching with large illumination changes. Experimental analysis of homologous and heterogeneous optical remote sensing images demonstrated the superior stability and capability of the proposed algorithm over commonly used algorithms, including the radiation-invariant feature transform, adaptive binning SIFT, gradient orientation modification SIFT, and LSS algorithms.

Automated summary

This study proposes a robust descriptor construction algorithm to improve the matching problem in heterogeneous optical remote sensing images by considering the internal gray value changes of ground objects. Experimental results demonstrate that the proposed algorithm exhibits better stability and capability in matching homologous and heterogeneous images compared to other commonly used algorithms.