Polynomial intensity correction for multimodal image registration

Based on the standard sum of squared differences (SSD) criterion, we develop a fast and robust multimodal image registration algorithm that incorporates a polynomial model to capture the complex relationship between the intensity values of two images. The resulting energy function can be efficiently optimized using the Gauss-Newton method in either joint or constrained fashion. Comparisons using simulated data reveal that our method has a similar capture range to that of the mutual information. Furthermore, we demonstrate accurate MR-PET registration results for images with abnormal structures.