Real-time registration of remote sensing images with a Markov chain model

The automatic registration of different remote sensing images of the same scene is an important problem in the processing of remote sensing images. Such images usually have different resolutions and computing an accurate registration of them in real time is thus often a challenging task. In this paper, a new statistical approach is developed for the accurate and efficient registration of two remote sensing images with different resolutions. The proposed approach utilizes a statistical model to evaluate the probability for each possible mapping between the two images and computes the one with the maximum probability for registration. Similar to most of the state-of-the-art methods for remote sensing image registration, the proposed approach assumes the existence of an affine transformation between the images to be registered. The registration is performed in three stages. In the first stage, the pixels in each image are efficiently partitioned into two sets with the Otsu's algorithm and four approximate equations for the parameters in the affine transformation are established. In the second stage, pixels in both images with significant edge geometric features are selected for potential pixel matching. In the third stage, based on the four approximate equations generated in the first stage, an Markov Chain Model-based method is used to efficiently compute the matching that can pair the selected pixels with the maximum likelihood. The parameters of the affine transformation can be determined from the pixel pairs with a least square regression approach. Experimental results on a number of pairs of remote sensing images show that this approach can generate registration results more accurate than those obtained with a few state-of-the-art approaches. In addition, real-time evaluation also shows that the approach is computationally efficient and can be used in real-time applications for remote sensing image processing.