An E cient Motion Estimator with Application to MedicalImage Registration

Image registration is a very important problem in computer vision and medical image processing. Numerous algorithms for registering multi-modal image data have been reported in these areas. Robustness as well as computational e ciency are prime factors of importance in image data registration. In this paper, a robust and e cient algorithm for estimating the transformation between two image data sets is presented. Estimating the registration between two image data sets is formulated as a motion estimation problem. We use an optical ow motion model which allows for both global as well as local motion between the data sets. In this hierarchical motion model, we represent the ow eld with a B-spline basis which implicitly incorporates smoothness constraints on the eld. In computing the motion, we minimize the expectation of the squared di erences energy function numerically via a modi ed Newton iteration scheme. The main idea in the modi ed Newton method is that we precompute the Hessian of the energy function at the optimum without explicitly knowing the optimum. This idea is used for both global and local motion estimation in the hierarchical motion model. We present examples of motion estimation on synthetic and real data and compare the performance of our algorithm with that of competing ones.