POWRE: Multiresolution Maximization of Mutual Information for Remote-Sensing
Morgan State University, Baltimore MD
Investigators
Abstract
0074929 Cole-Rhodes This project will address the problem of registering images produced by remote sensing satellite missions. Large amounts of data are generated by these systems from multiple-time observations of the same features, or from many sensors mounted on the platform which may simultaneously observe the same features. The data produced may be in different spectral ranges and/or at different spatial resolutions. All this data must be integrated in order to provide a better understanding of earth and space phenomena, and image registration is the first step in this process. Image registration is the process which determines that transformation which provides the best match between a reference image and an input image. The goal is to define fast, accurate methods for the automatic registration of multisensor remotely sensed data. The optimal match can be determined using some similarity measure, such as correlation or the Hausdorff distance between two images. This project will apply a different similarity measure from Information theory, called Mutual Information, to the registration of remote-sensing images. It will investigate whether mutual information can be successfully applied to cross-register images produced by different remote-sensing sensors. The most basic method of finding the optimal transforma-tion, is by an exhaustive search over the space of allowable transformations. This method is computationally expensive and not very efficient. The project will investigate other search strategies based on the use of mutual information, while also attempting to speed up the search procedure by the use of a multiresolution strategy. This strategy will be based on the wavelet decomposition, since this has already been successfully applied to automate the image registration process for remote sensing data. This project invites a new direction of research for the Principal investigator, and it is directly integrated into the developing graduate program in the School of Engineering at Morgan State University. ***
View original record on NSF Award Search →