HyperMap Spectral Imagery Registration Software uses an information-based approach to compute the registration between spectral images that maximizes the measure of mutual information between them. This differs from traditional feature-based approaches in that it uses all of the available image data and is robust to differences encountered when working with multimodal images, such as nonlinear photometric transformations and contrast inversions. To speed up processing time, the optimization works over an image scale space from low resolution to full resolution, as well as a transformation parameter space from translation-only to a full-projective transform. A second non-rigid registration stage can optionally be enabled that further maximizes mutual information by refining the initial rigid registration solution using a free-form deformation model. This is useful in cases where the misalignment has more degrees of freedom than a rigid image transformation can compensate. Many of the operations have been parallelized for use on modern GPU hardware.