Research Of Medical Image Registration Based On Total Variational Constraints

In the medical field,doctors often register multiple images of the same patient in multiple modes or multiple imaging results in the same mode.This process can be attributed to the inverse problem in the field of mathematical physics,and most of the inverse problems are ill-posed.Since the introduction of the concept of well-posed by Hadamard in 1923,people have more and more researches on inverse problems,and the theoretical methods for solving inverse problems have also received more and more attention.Under such a background,regularization theory has entered an unprecedented development climax.This paper discusses the problem of nonlinear ill-posed problems.Firstly,we show the advantages,disadvantages and limitations of the existing regularization theory.We consider that the Total Variation(TV)regularization method can better deal with the discontinuity of the function solution,and can be used to preserve border information more clearly when applied to images.At the same time,combining the homotopy perturbation method's characteristics which do not depend on the initial value of the function and has the characteristics of large-range convergence,an iterative format based on the total variational constraint is constructed.Secondly,the condition that the iterative format converges to the minimum value of the target functional are given.Proving that the error of the iterative sequence under the corresponding conditions is degressive,and that the constructed iterative format converges in both cases with data disturbance and without data disturbance.Finally,for the medical image registration problem,the numerical simulation is carried out using the method proposed in this paper.Two different medical image registrations were used as the background,and the brain and lung image registration are respectively considered.Comparing the constructed algorithm with the sparsity-constrained method.The results show that on the premise of ensuring the same number of iteration steps,the iterative format constructed in this paper can achieve a higher correlation coefficient between the registration result and the reference image in a shorter period of time,ie achieving a better registration effect.The advantages of the iterative format proposed in this paper are high efficiency and high accuracy.