Study Of Image Registration Based On Improved Demons Algorithm

Demons algorithm attains extensive attention in image registration because of its effectiveness and simplicity.According to different demands,scholars have come up with plenty of improved algorithm,such as Demons algorithm,and Diffeomorphic Demons algorithm.The Active Demons algorithm adjusts the convergence speed and registration accuracy by homogenizing coefficients and negative driving forces,and overcome the defect that Demons algorithm can not effectively register the image with relatively large deformation.But in the process of registration,the pixels of Active Demons algorithm move in a free state,which destroys the topology of the original image.The advantage of Diffeomorphic Demons algorithm is that the concept of Lie group is introduced,which keeps the original topology in the registration process.In theory,the Diffeomorphic Demons algorithm can be applied to all kinds of deformation image registration,but due to the problem that the calculation can not completely maintain the original image topology problem,making the algorithm has too much computation,poor adaptability and so on.In this paper,we focus on the Active Demons algorithm and the Diffeomorphic Demons algorithm,and give the corresponding improvement based on the physical essence of the image registration problem and the appropriate mathematical theory.Aiming at the problem that convergence speed and registration precision of Active Demons algorithm are not uniform,this paper proposes an improved algorithm � F-angle Demons algorithm.Through the analysis of the principle of Active Demons algorithm and the effect of registration,it is found that the main reason for the existence of the registration error of Active Demons algorithm is that the gradient of the floating point M and the reference image F is sometimes too big.This problem leads directly to the deviation of the driving force from the correct direction,resulting in registration errors.Therefore,the introduction coefficient f(?) increases the influence of the negative driving force at the gradient acute angle,and reduces the negative driving force hindering effect at the gradient obtuse angle.Numerical experiments show the feasibility of the algorithm.Finally,the control coefficients are introduced into the driving force equation ? to improve the accuracy of the registration algorithm under the condition of high convergence speed and validated by numerical experiments.For poor adjustment of Diffeomorphic Demons algorithm,this paper proposes an improved algorithm based on gradient information Diffeomorphic Demons algorithm.In particular,it introduces the energy function gradient information to further define the error term update field u,and deduced by rigorous mathematical calculation formula.Multiple sets of numerical experiments show that the proposed method has less error in internal calculation and high adaptability of the algorithm.