Multi-sensor Remote Sensing Image Registration by Maximization of Arithmetic-Geometric Mean Divergence
最大化算术-几何均值距离的多传感器遥感图像配准

Mutual information is an important method for multimodal medical image registration. It measures Kullback-Leibler (KL) divergence between two probability distributions. The connection between KL divergence and Shannon inequality is investigated. Base on the connection, a novel measure, Arithmetic-Geometric (AG) mean divergence, is proposed. It can be used for alignment of remote sensing images acquired by different sensors. Unlike KL divergence, the new measure is symmetry and do not require the condition of absolute continuity to be satisfied by the probability distribution involved. AG divergence measure is applied to one-dimensional simulated signals, and to affine registration of Thematic Mapper (TM), Satellite Positioning and Tracking (SPOT) and Synthetic Aperture Radar (SAR) remote sensing images. The performance of AG divergence measure is validated by experiments. The results show that AG measure do not require the approximate linear relation of pixel intensity value in image pairs, and is practicable even though the gray values of images are much different from each other.