Registration of image cubes using multivariate mutual information

A new method for simultaneously registering a collection of multispectral or hyperspectral images (spectral image cubes) using mutual information is presented. In this paper, we derive a new algorithm based on mutual optimization between pairs of images and extend it to any finite number of images. We present results on the convergence and stability of the new algorithm, showing that the algorithm registers each image to a weighted average of its alignment with every other image. This method has application to other multiimage applications as well.