Registration of multiple regions derived from the optical flow model and the active contour framework

In this paper, we present a new paradigm to carry out the nonrigid registration of multiple regions with a dense deformation field derived from the optical flow model and the active contour framework. The method can merge different tasks such as registration, segmentation and incorporation of prior knowledge into a single framework. The technique is based on finding the deformation field which minimizes an energy functional designed to be minimal when the curves selected in the moving image have reached the target objects. This way, we obtain registration results that are accurate on the considered contours with a deformation field smooth over the whole image. Finally, the registration of multiple regions is handled with a label function. A well-suited application of our registration model is the atlas-based segmentation in medical imaging. Results on 2D synthetic and real data show the performance of our registration algorithm.