A New Curves-Based Local Image Registration Method

In this paper, we propose a new curve-based local image registration method, which can elastically register the images under the guide of feature curve. The feature curves extracted from the images are modeled by B-splines. The action domain of each feature is divided into two parts: inner action domain and outer action domain. In inner action domain, local coordinate system of feature curve-pair is built, and the points are transformed according to the corresponding relationship of curve-pairs. In outer action domain, some point-pairs are sampled as evenly as possible by the correspondence between inner boundary and outer boundary. And then the outer transformation function is obtained by interpolating these point-pairs. For the frequent computation of the shortest distance, restricting region and together computation are adopted to improve the efficiency of algorithm. The experimental results show that our algorithm is a reasonable and feasible method and has great efficiency in elastic image registration based on feature curves.