Fast registration of diffusion tensor image based on polynomial expansion

An improved fast registration algorithm of diffusion tensor images is proposed based on polynomial expansion. The edge feature points of an enhanced image of anisotropy measure are extracted by Laplacian operator. An image transformation model is given, in which the characteristics of the image are defined by the edge feature points and the direction of the tensor image is corrected therewith. By use of the registration model based on polynomial expansion, a global dense registration of pixel points is done for the image of anisotropy measure, the direction of which has already been corrected. The experimental results show that the method can effectively improve the registration efficiency and registration accuracy.