Abstract: A structure compensation-based registration method was proposed for chest multimodal magnetic resonance imaging(MRI)images with mismatched structures.First of all,floating image was pre-registered to reference image in similarity transformation.Then the pre-registration image was compensated by segmenting and morphological processing methods,a floating template,whose tissue structures matched with the reference image’s structures,was generated.After that,the floating template was non-rigidly registered to the reference image,the deformation field was extracted from the registration.Finally,the deformation field was applied to the pre-registration image,the registered image was generated.The Jaccard index of the mismatched structures was used as measurement,frame models and clinical images were used to validate the proposed method.Results indicate that the proposed method is more effective than the traditional mutual information based method in chest multimodal MRI images with mismatched structures.