Non-rigid Medical Image Registration And Application Based On Sparse Representation

Medical image registration is an important part of computer vision and medical image processing.It can help doctors judge and treat diseases more effectively.It also has important significance for medical image reconstruction and fusion.However,the registration process of medical images is complicated,and there are irregular deformations among most of the registered images,and the medical images is lower than natural image and contain noise,gray-scale shifting fields,etc.There are still many problems at present.Therefore,it is urgent to explore new registration algorithms with good robustness and wide application range.This article takes non-rigid brain medical images from Brainweb,Harvard Medical School,and other standard libraries as the research object,and divides the imaging devices into single-mode and multi-modal types and conducts research.The The main research contents are as follows:(1)This paper proposed the non-rigid medical image registration method of multilayer P-spline and sparse coding according to two problems,the first one is the mismatch because the traditional similarity measure is easily affected by the gray bias field and the second problem is that single-layer P-spline transformation model is unable to accurately select the initialized grid density.This method uses sparse coding as a similarity measure and multilayer P-spline as a geometric transformation model to simulate non-rigid geometric deformation.The proposed method can accurately select the mesh density and effectively overcome the effect of the gray-shift field on registration,reduce the root-mean-square error,and improve the accuracy and robustness of registration.(2)For sparse coding similarity measure,it is robust to gray-scale migration in non-rigid medical image registration,but it is only applicable to the problem domain of single-modality medical image registration.So I proposed a non-rigid multimodal medical image registration method based on multi-channel sparse coding.This method treats the multimodal registration problem as a multi-channel registration problem.Each mode after dividing the channel is operated in a separate channel,and then a weighted sum of each channel is performed.The proposed method can effectively overcome the effect of gray shift field on registration in multimodal medical image registration and improve the accuracy and robustness of registration.(3)Based on the two registration methods proposed above,the entire non-rigid medical image registration system was designed and developed using Matlab.Single-mode,multi-mode non-rigid medical image registration and multi-modality medical image fusion functions are realized.