Comparison of Popular Non-Rigid Image Registration Techniques and a New Hybrid Mutual Information-Based Fluid Algorithm

Recently there has emerged a need to compute multimodal non-rigid registrations in a lot of clinical applications. To date, the viscous fluid algorithm is perhaps the most adept method at recovering large local misregistrations that exist between two images. However, this model can only be used on images from the same modality as it assumes similar intensity values between images. This paper presents a solution to this problem by proposing a hybrid non-rigid registration using the viscous fluid algorithm and mutual information (MI). The MI is incorporated via the use of a block matching procedure to generate a sparse deformation field which drives the viscous fluid algorithm. This algorithm is compared to two other popular local registration approaches, namely Gaussian convolution and the thin-plate spline warp. Results show that the thin-plate spline warp and the MI-Fluid approach produce comparable results. However, Gaussian convolution is the superior choice, especially in controlled environments.