Application of Algorithm with Novel Successive Approximate Registration Method of Medical Image

Abstract

Medical image registration is widely used in the clinical field. A simple and effective registration method for medical images is proposed, which is not only used images’ hierarchical information, but also the successive approximate method is used. It mainly includes two steps, which are the coarse registration and the fine registration. In the coarse registration procedure, the couple images’ edge features are extracted. Further more, the contour line features are extracted based on the edge features. The contour lines are selected as registration objects to accomplish the coarse registration based on the Principal Axis method. In the second registration procedure, i.e. the fine registration, the hierarchical information of the images are obtained by the convolution operator and pooling method firstly. Then these information are matched from the last scale to the first scale to obtain the last registration result. The proposed method’s validity is proved by different registration experiments, and the proposed method can also be applied to incomplete images’ registration. The error of registration is less than one pixel unit in the different registration experiments of the proposed method.