For the problem of image registration based on mutual information ignore global spatial information of the two images, and the lungs PET image is so fuzzy that has fewer correlations with the lungs CT image. This paper proposed a coarse-to-fine image registration method based on region information and mutual information. Experimental results show that this method can effectively to achieve lung image registration which globally aligns the chest region in PET and CT images and locally overlaps nodule lesion anatomical structure region in CT image and nodule highlighted functional area in PET image.