Magnetic Resonance Imaging Fusion by 3D Compactly Supported Shearlet Transform

T2* and quantitative susceptibility mapping (QSM) of magnetic resonance imaging (MRI) images provide different type inner structure information of scanned organs. If they can be properly fused into one set, the details of the scaned organ can be revealed more clearly. In this paper, a 3D MRI image fusion method based on 3D compactly supported shearlet transform (3D-CSST) and 3D dual tree compactly supported shearlet transform (3D-DT-CSST), is proposed, which can overcome the limitation, loss of inter layer correlative information, of conventional 2D image fusion methods. 3D-DT-CSST is our modification of 3D-CSST, which is approximate shift invariant. It can improve the performance of fusion method. The proposed method is evaluated by 4 groups of MRI images of human brains. The results suggest that the proposed method has a better performance than conventional 2D wavelet, 2D DT-CWT and 3D wavelet, 3D DT-CWT based fusion methods, and 3D-DT-CSST based method is better than 3D-CSST based method.