Nonlinear intensity difference correlation for multi-temporal remote sensing images

Compared to geometric distortions exhibited in remote sensing images, radiometric deformations are less addressed in the literature and linear variations are usually assumed during image matching or registration. This paper proposes a novel robust and automatic image matching approach for multi-temporal and multi-sensor remote sensing images which usually present non-linear radiometric changes. We introduce a non-linear intensity difference correlation (NIDC) algorithm that aims to reduce the impacts of non-linear intensity differences during image matching. Differences of illumination intensity are accounted for and modeled in the NIDC algorithm through the extension of traditional feature-based matching techniques. Our proposed approach has been tested with typical multi-temporal and multi-sensor remote sensing images characterized by either dense or sparse ground control points (GCPs). Experimental results demonstrate that our matching approach outperforms common image matching techniques such as cross correlation (CC), scale-invariant feature transform (SIFT), and speeded-up robust features (SURF) with respects to matching success rate and robustness in test data.

► We propose a novel non-linear intensity difference correlation (NIDC) approach for remote sensing image matching.
► Non-linear illumination variations are modeled through the extension of traditional feature-based matching techniques.
► Experiments demonstrate that our matching approach outperforms several traditional image-matching techniques.