Image fusion involves merging two or more images into a single image with more details and information. Recently, multiscale transform approaches are very useful for analyzing the information content of images for the purpose of fusion. The nonsubsampled contourlet transform (NSCT) can provide a flexible multiscale, multidirection, and shift-invariant image decomposition for image. In order to compose the coefficients effectively in different subbands, spatial frequency is brought forward to select the adapted coefficients. Experimental results demonstrate the proposed fusion algorithm can improve the resultant image well, both in visual effect and objective evaluation criteria, and the method is robust when applied in multifocus or multisensor images.