Abstract: An image fusion algorithm in nonsubsampled Contourlet domain is presented based on Human Visual System (HVS) and source image characteristics. Moreover, the influence on fusion result of decomposition levels and directional decomposition numbers is discussed. Closed Loop Feedback (CLF) is introduced into low-pass subbands to obtain optimal fused weights adaptively. In high-pass subbands, Activity Measure (AM) is defined based on region energy to enhance contrast of fused images and protect detail information of source images. Experiment results show that the proposed fusion technique is robust and the fusion images have ideal clear and continue edges.