ABSTRACT
Probably one of the most popular methods for deformable image registration (DIR) is the so-called “Demons” method, initially proposed by Thirion (1996, 1998). DIR between images I1 and I2 can be considered as the minimization of an energy function, representing a tradeoff between image similarity and deformation regularity. In the Demons algorithm, similarity and regularity are optimized in consecutive but independent operations. The algorithm consists of an iterative procedure composed of two steps. The first step aims at defining an active force at each voxel. This force is directed in the opposite direction of the image gradient, with a magnitude proportional to the gray-level difference between the two images. The second step is a Gaussian smoothing of the resulting vector field. This two-step procedure is iterated until convergence.
